Neural Correlates of Eureka Moment

Intelligence 62 (2017) 99–118

Contents lists available at ScienceDirect

Intelligence

journal homepage: www.elsevier.com/locate/intell

Neural correlates of Eureka moment

Giulia Sprugnoli a, Simone Rossi a, Alexandra Emmendorfer b, Alessandro Rossi a, Sook-Lei Liew c, Elisa Tatti a, Giorgio di Lorenzo e,f, Alvaro Pascual-Leone b, Emiliano Santarnecchi a,b,d,⁎ a Brain Investigation & Neuromodulation Laboratory, Department of Medicine, Surgery and Neuroscience, University of Siena, Siena 53100, Italy b Berenson-Allen Center for Non-Invasive Brain Stimulation and Division of Cognitive Neurology, Beth Israel Medical Center, Harvard Medical School, Boston, MA 02120, USA c Chan Division of Occupational Science and Occupational Therapy, University of Southern California, Los Angeles, CA, USA d Center for Complex System Study, Engineering and Mathematics Department, University of Siena, Siena 53100, Italy e Laboratory of Psychophysiology, Chair of Psychiatry, Department of Systems Medicine, University of Rome “Tor Vergata”, Rome, Italy f Psychiatry and Clinical Psychology Unit, Department of Neurosciences, Fondazione Policlinico “Tor Vergata”, Rome, Italy article info abstract

Article history: Insight processes that peak in “unpredictable moments of exceptional thinking” are often referred to as Aha! or Received 9 September 2016 Eureka moments. During insight, connections between previously unrelated concepts are made and new pat- Received in revised form 5 March 2017 terns arise at the perceptual level while new solutions to apparently insolvable problems suddenly emerge to Accepted 5 March 2017 consciousness. Given its unpredictable nature, the deﬁnition, and behavioral and neurophysiological measure- Available online 28 March 2017 ment of insight problem solving represent a major challenge in contemporary cognitive neuroscience. Numerous

Keywords: attempts have been made, yet results show limited consistency across experimental approaches. Here we pro- Insight vide a comprehensive overview of available neuroscience of insight, including: i) a discussion about the theoret- Eureka ical deﬁnition of insight and an overview of the most widely accepted theoretical models, including those Aha debating its relationship with creativity and intelligence; ii) an overview of available tasks used to investigate in- Cognition sight; iii) an ad-hoc quantitative meta-analysis of functional magnetic resonance imaging studies investigating fMRI the Eureka moment, using activation likelihood estimation maps; iv) a review of electroencephalographic evi- EEG dence in the time and frequency domains, as well as v) an overview of the application of non-invasive brain stim- ERPs ulation techniques to causally assess the neurobiological basis of insight as well as enhance insight-related Non-invasive brain stimulation cognition. Neuroenhancement NIBS © 2017 Elsevier Inc. All rights reserved. Creativity

Contents

1. Introduction...... 100 2. Defining the topic: definitions,theories,andtasks...... 100 2.1. Aninsightintoinsight...... 100 2.2. Definingprocesses...... 101 2.3. Insightorinsights?...... 102 2.4. Insighttasks...... 102 2.5. Creativityandinsight...... 103 2.6. Multifactorial nature of Aha!:intelligence,memory,attention,mood,cognitivecontrolandsleep...... 103 2.6.1. Intelligenceandmemory...... 104 2.6.2. Attention...... 104 2.6.3. Mood...... 104 2.6.4. Sleepandcognitivecontrol...... 104

⁎ Corresponding author at: Berenson-Allen Center for Non-Invasive Brain Stimulation and Division of Cognitive Neurology, Beth Israel Medical Center, Harvard Medical School, Boston, MA, USA. E-mail addresses: [email protected] (G. Sprugnoli), [email protected] (S. Rossi), [email protected] (A. Emmendorfer), [email protected] (A. Rossi), [email protected] (S.-L. Liew), [email protected] (E. Tatti), [email protected] (G. di Lorenzo), [email protected] (A. Pascual-Leone), [email protected] (E. Santarnecchi).

2.7. Overviewofneurophysiologicalevidenceaboutinsightprocesses...... 104 3. Methods...... 104 3.1. Literaturesearch...... 104 3.2. fMRIstudies...... 106 3.2.1. ComputationoftheALEmeta-analysismaps...... 106 3.3. EEGstudies...... 106 3.4. Non-invasivebrainstimulationstudies...... 107 4. Results...... 107 4.1. Insight-relatedfMRIactivations...... 107 4.2. Insight-relatedERPs...... 108 4.3. Brainoscillationsduringinsight...... 109 4.4. Brainoscillationsduringmentalpreparationandresting-stateactivity...... 111 4.5. NiBSandinsight...... 111 5. Discussion...... 112 5.1. Theinsightnetwork...... 112 5.2. The timing of an insightful solution: event-related potentials during Eureka! ...... 113 5.2.1. Negativepotentialsaround400ms...... 113 5.2.2. Laternegativecomponents...... 113 5.2.3. Positivecomponentsaround300ms...... 114 5.2.4. Laterpositivecomponents...... 114 5.3. Hemisphericasymmetryandinsightfulprocessing...... 114 5.4. Electricsignaturesofaninsightfulmind...... 114 5.5. InsightandNiBS...... 115 5.6. Caveatsandfuturedirections...... 115 6. Conclusion...... 115 Financialdisclosures...... 116 Acknowledgements...... 116 AppendixA. Supplementarydata...... 116 References...... 116

1. Introduction principle of displacement while taking a bath, reportedly ran naked down the street shouting “Eureka!”. This funny anecdote highlights Although research on insight processes began over a century ago the unpredicted, unfettered nature of Aha! moments, thought of as “a with Köhler's observations on the problem-solving abilities of chimpan- special gift of Muses” by the Greek. While solid theory has been pro- zees (Kohler, 1925), a comprehensive definition of “insight processes” posed for the biological network of intelligence (Jung & Haier, 2007), remains elusive. During the last twenty years, several theories have valid scientific explanations are largely lacking for insight. This leaves been proposed to explain the insight phenomenon. Over the past de- the Eureka moment as one of the most intriguing and unexplained pro- cade, experimental support for some of these theories has been gathered cesses of the human mind (Sternberg & Davidson, 1995), despite many thanks to recent advances in neuroimaging and neurophysiological relevant correlations with fluid intelligence (Paulewicz, Chudersky, & techniques. In the present review, we provide a comprehensive summa- Necka, 2007; Sternberg & Davidson, 1995), switching ability and work- ry of the neuroscience of insight. We first provide an overview of the ingmemory(WM)capacity(Murray & Byrne, 2005). It was not until the most relevant theoretical definitions and the most commonly used beginning of the 20th century that Gestalt psychologists attempted to tools for the investigation of insight moments. Second, we present orig- create a proper definition of insight (Dietrich & Kanso, 2010), describing inal results from a quantitative meta-analysis of available functional it as “a process based on reconstructing the core of a problem, rethink- magnetic resonance imaging (fMRI) data, as well as a summary of the ing its basic assumptions and originating a new and creative solution, a evidence collected with electroencephalography (EEG), focusing on process usually occurring in an unexpected and unpredictable manner” brain oscillations and event-related analysis. Third, we critically discuss (Kohler, 1925). emerging evidence from perturbation-based and neuromodulatory ap- To better characterize the Aha! moment, a valid heuristic approach proaches, such as repetitive transcranial magnetic stimulation (rTMS) might be to discard what is not considered insight problem-solving. In and transcranial electrical stimulation (tES), which add a causal dimen- general, problem-solving strategies can be divided into three types: an- sion to traditional neuroimaging mapping data, allowing for the tran- alytical problem-solving, memory retrieval, and insight (Novick & sient modification of regional brain dynamics underlying insight Sherman, 2003). Analytical problem-solving is characterized by three processes. Finally, we address the possibility of using non-invasive main features: (i) it is deliberate and predominantly conscious, (ii) it ad- neuromodulation as a tool to enhance insight problem-solving abilities. vances step by step from the initial processing of information to the resolution and (iii) its steps are available to WM, so that subjects are able to 2. Defining the topic: definitions, theories, and tasks explain in details how they were able to approach the solution. In contrast to analytical problem-solving, which is marked by a deep under- 2.1. An insight into insight standing of the problem, memory retrieval processes can be described as a simple mental retrieval of previously acquired knowledge, which Many great scientific discoveries have relied on insight moments fits to the problem at hand (Aziz-Zadeh, Kaplan, & Iacoboni, 2009). (e.g., Newton's finding of the law of gravitation, Kekulé's discovery of Insight problem-solving is thought to be very different from these the structure of benzene, Poincaré's discoveries in mathematics, other two strategies. The Aha! moment consists of a sudden, unexpect- Einstein's first theorization of the General Relativity theory; ed, and somehow “obvious” solution that cannot be explained by a se- Sandkühler & Bhattacharya, 2008). The first known Aha! moment typi- quential solution process. Unlike analytic problem-solving, the cally refers to Archimedes of Syracuse, who, after discovering the subjects cannot readily explain the exact path they followed to reach G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 101 the solution. Often, a sense of being stuck precedes the insight phenom- Multiple theories have been proposed to explain the Aha! moment, in- enon, and the way out of this mental impasse is provided by the creation cluding the Progress Monitoring Theory (MacGregor, Ormerod, & of novel associations, rendering insight problem-solving a multistep Chronicle, 2001), the Representational Change Theory (Knoblich, process (Bowden & Jung-Beeman, 2007; Bowden, Jung-Beeman, Fleck, Ohlsson, Haider, & Rhenius, 1999) and a recent conceptualization by & Kounios, 2005). Interestingly, insight solutions are generally more ac- Bowden and Beeman (Bowden et al., 2005). curate than analytical ones (Salvi, Bricolo, Kounios, Bowden, & Beeman, The Progress Monitoring Theory by MacGregor, Ormerod, and Chron- 2016). A possible explanation could rely on the all-or-none nature of icle, is based on the hill-climbing method, and has been adapted to the Aha! moment, which does not allow the subject to provide intermediate performance of a classic insight problem, the Nine-dot problem responses and contributes to the positive burst of emotion when an in- (Fig. 1). The hill-climbing method can be applied to problems that sight comes to consciousness (Salvi et al., 2016). have many solutions. It begins with the assumption of a random solu- Recently, a new definition of the Eureka moment (Kounios & tion, which is subsequently manipulated making small changes, each Beeman, 2014) described it as “any sudden comprehension, realization, time getting closer to the goal. When such a process does not produce or problem solution that involves a reorganization of the elements of a results anymore, the subject comes to an impasse and starts searching person's mental representation of a stimulus, situation or event to yield for a new approach, following a trial and error heuristic path. This theo- a nonobvious or nondominant interpretation”.Inthisview,theprocess ry implies that solvers constantly monitor their own progress in order to is happening almost entirely at the unconscious level. Interestingly, a promptly switch to a different problem-solving strategy in case the cur- mental impasse is not considered a necessary feature of insight in this rent one is not successful. This theory suggests that the Aha! moment definition, because this aspect would implicitly exclude insight mo- may be achieved with an incremental approach, with constant monitor- ments that occur when: (i) subjects are not concentrating on problem ing of the own cognitive processes as a pivotal feature, making the Eure- solutions (e.g. while taking a shower), (ii) use the analytic problem- ka moment more like a conscious epiphenomenon of a general solving method but have not yet reached an impasse, or (iii) are sud- problem-solving process (Theory of Business as Usual; Bowden et al., denly struck with a new idea. Finally, the authors considered the posi- 2005) rather than a burst of uncommon cognitive processes (Theory of tive burst of emotion (that usually accompanies the Aha! moment) as the Special Process; Bowden et al., 2005; see also the paragraph an additional, but not necessary, feature, as it is not always present. “Multifactorial nature of Aha!: intelligence, memory, attention, mood, In summary, a few features of the Aha! moment seem to systemati- cognitive control and sleep”). cally overlap across theories and models: (i) insight takes form as a sud- In contrast to the Progress Monitoring Theory, Knoblich and col- den comprehension of the solution to a problem; (ii) the subject has no leagues introduced the Representational Change Theory (Knoblich et al., access to the analytical steps leading to the insight moment; (iii) insight 1999) to emphasize the importance of the reorganization of a problem's triggers a sense of surprise and a burst of positive emotions spanning representation. In general, the authors suggest that the main issue of from satisfaction to euphoria. problem-solving relies on the spontaneous human tendency to set unnecessary constraints through a very restricted and reductive mental 2.2. Defining processes representation of the problem at hand, which is modeled on past and consolidated knowledge. To overcome the impasse and reach the solu- A proper understanding of insight problem-solving requires going tion, subjects must: (i) remove these unnecessary constraints by beyond definitions and considering the underlying mental operations. deactivating the recalled knowledge linked to the problem, and (ii)

Fig. 1. Insight tasks. Examples of different tasks used to assess insight problem-solving are shown, highlighting the heterogeneity of available assessment tools reported in the literature. Pie chart refers to the percentage of studies implementing each different task (“Chinese tasks” referred to Chinese logogriphs, characters, character-generation task, Chinese riddle and puzzles; “others” referred to: ambiguous sentences, technical problems, prototypes, number reduction task, brainteasers; see the paragraph “Insight tasks” for additional details and also Tables 1, 2, S1, S2, S3, S4). 102 G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 make a decomposition of the elements of the task by dividing it into per- in general may not be appropriate before extensive experimental EEG ceptual chunks. Constraints and chunks might differ: local and global and fMRI validations. constraints respectively apply to a single part or to the whole problem Despite the various conceptualizations discussed above, a clear the- representation, while loose and tight chunks can or cannot be divided oretical representation of insight is still not available. Except for core el- into further chunks. Knoblich and colleagues ascribe to the mastering ements, such as the creation of mental associations and the of these mechanical and fairly analytical steps the ability to create a restructuring of the problem in a novel and useful way, more robust new representation of the problem and reach the Aha! moment, some- neurophysiological models of the Eureka moment are necessary to settle how again minimizing the unconscious nature of the insight on an evidence-based definition. phenomenon. Both the Progress Monitoring Theory and the Representational Change 2.3. Insight or insights? Theory have received some experimental support, with the former showing a better fit to multi-step problems (e.g. Nine-dot problem, A topic of lively discussion is whether the Eureka process is specific see Fig. 1), and the latter being more suited to single-step problems to the problem-solving domain, and to what extent it is separated (e.g. Matchstick arithmetic problems, Fig. 1). from the analytical method. Although the majority of scientists consider Bowden and Beeman have proposed another theory to explain the the insight process as related only to higher order problem-solving, Eureka moment. They suggest that the steps leading to an insight mo- Bowden and colleagues suggest that it could be a more general phe- ment include (i) a strong activation of useless consolidated information, nomenon encompassing perception (e.g. the sudden recognition of an coupled with a weak (unconscious) activation of new information cru- object in a blurred or ambiguous picture) and language comprehension cial to reach the solutions; (ii) a secondary integration and reorganiza- (e.g. the sudden comprehension of a joke or metaphor; Bowden & tion of the elements in a non-dominant way respect to the initial Jung-Beeman, 2007). According to this view, the Aha! moment reflects representation of the problem, and finally (iii) the restructured repre- a generalizable process of acquired knowledge across domains, includ- sentation reaching consciousness (Bowden et al., 2005). The “uncon- ing perception, language, decision making, problem-solving, etc. Thus, scious trajectory” was the principal limitation in investigating the all moments of insight require finding a particular piece of information neural basis of insight process in humans. However, the intrinsic diffi- that has not been declared from the background, and integrating it with culty in measuring – or quantifying – a construct cannot constitute a prior knowledge at the unconscious level, to reach a conscious solution limitation to our understanding of human cognition and has not been (Bowden et al., 2005). Moreover, they postulate a discrete overlay be- a constraint for other controversial topics, such as creativity and mem- tween the insight problem-solving processes and non-insight process- ory retrieval (Schooler & Melcher, 1997). More appropriate tools are es, with the former implying more unique features than the others. needed in order to reveal the underlying anatomy and physiology of in- Taking into account the various fields where insight-like processes sight problem-solving, with fMRI, EEG and non-invasive brain stimula- seem to be involved, Bowden and colleagues postulate that all types of tion (NiBS), and combinations of these methods, constituting valuable Eureka moments would need a basic neural network and mental opera- solutions. Neurophysiological correlates of insight supporting this tions, which are partially shared with analytical problem-solving framework through spatially and temporally distributed inter- (Schooler & Melcher, 1997). If so, it might be possible to identify a com- hemispheric activations are now emerging as its first unconscious step mon insight process across all domains – as partially revealed for math- (shared also with creativity process, see also “Creativity and insight” ematical and verbal tasks (Bermejo, Castejon, & Sternberg, 1996) – and paragraph; Simonton, 2001; Schooler & Melcher, 1997). An initial specific insight processes for each subfield of cognition where Eureka weak semantic activation, characterized as the first stage of problem- moments might occur (Bowden & Jung-Beeman, 2007). solving, occurs in the right hemisphere as a general semantic encoding process, while the fine semantic encoding takes place in the left hemi- 2.4. Insight tasks sphere (Bowden et al., 2005). The weak coarse semantic activation is as- sumed to be crucial in insight problem-solving because it implies the Given the multitude of non-univocal theories and definitions of in- access of alternative meanings and distant semantic associations be- sight, many tasks have been developed to address it. Gestalt psycholo- tween the items at hand. Considering it is a faint and non-dominant ac- gists were the first to introduce insight tasks almost a century ago, tivation, it is initially blocked and suppressed by potentially misleading such as the “Nine-dot problem”,the“Dunker candle task” and the left-sided activations, leading to an impasse. When subjects overcome “Eight Coin problem” (for a review see Chu & MacGregor, 2011 and the consolidated and dominant coding, an Aha! moment is likely to Fig. 1). Despite being widely used in the assessment of insight occur. To verify the hypothesis of an essential role played by the right problem-solving, these classical insight problems have a number of lim- hemisphere, Bowden and colleagues conducted a series of experiments itations: (i) they are usually so difficult that only a small percentage of using a visual-hemifield presentation of Compound Remote Associa- participants can solve them in an experimentally compatible amount tions (CRA, Fig. 1) problems with a priming paradigm (Bowden & of time, (ii) their solution usually requires a very complicated Jung-Beeman, 1998). The results confirmed their prediction: subjects generative-operative process, (iii) they cannot be re-tested due to provided faster responses when stimuli relevant for the solution were their single-trial/item nature, (iv) they are mostly visuo-spatial prob- presented to the left visual field/right hemisphere rather than to the lems (independent from the verbal knowledge) and (v) very heteroge- right visual field/left hemisphere (see also Beeman & Bowden, 2000; neous. Moreover, these tasks are characterized by the fact that they Bowden & Jung-Beeman, 2003). The authors suggest hemispheric asym- cannot be solved through other problem-solving strategies (i.e., the an- metry (with the prominent role of right hemisphere) as the structural alytical method) and thus depend fully on insight problem-solving, lim- and functional basis for insight problem-solving. Ultimately, in line iting experimental controls by shifting strategies. Therefore, these with studies reporting the role of the right anterior superior temporal classical insight tasks are not ideal for appropriately controlled experi- gyrus in the creation of distant semantic relations (Mashal, Faust, & ments (Bowden et al., 2005). Hendler, 2005; Mason & Just, 2004), Bowden and colleagues hypothe- To overcome the limitations of classical insight tasks, a second gen- sized the unconscious reorganization of the problem as a result of the eration of insight problems was developed. Largely based on verbal activation of this region, consequently shifting the attention towards comprehension, they are easier to solve when used with the this brain area as a core node of the insight network in humans (the appropriate linguistic population, they include many more items for same region is also highlighted by our fMRI analysis, see fMRI discussion each trial type and they are generally faster to solve. Examples are section). However, it is worth noting that these findings are specificfor tasks based on riddles or anagrams, as well as ad-hoc designed tasks insight solutions achieved with CRA task, so their applications to insight such as the matchstick arithmetic problems, Chinese logogriphs, rebus G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 103 puzzles, the Remote Associates Task (RAT) and the CRA. For a graphical underscoring the innovative measurement of insight introduced by depiction of each task, see Fig. 1. the CRA task. The Rebus puzzles (MacGregor & Cunningham, 2008) are composed by visual and verbal information that has to be integrated to find a fa- 2.5. Creativity and insight miliar phrase. In this type of task, the subject must relax the constraints implemented by grammatical rules and reading (i.e., deactivate the Most classical insight tasks are also used to study creativity. It is recalled knowledge and assumptions linked to language) to bring out therefore important to address the relationship between creativity the real meaning of the compound stimulus (as required in the Repre- and insight. Whether insight moments should be considered a compo- sentational Change Theory). nent of human creativity remains controversial, with experimental The Matchstick arithmetic problems (developed by Knoblich in and theoretical work supporting conflicting views (Schooler & 1999) are false equations written in Roman numerals using matchsticks, Melcher, 1997). Indeed, creativity is one of the most complex human therefore they are not dependent on language and verbal comprehen- abilities, the primum movens of progress and innovation in all fields. Cre- sion. Subjects can move only one matchstick to transform the false ativity is broadly understood as the ability to change existing thinking statement into a correct mathematical relation (Knoblich et al., 1999). patterns, producing something that is useful, novel and generative The anagrams (Novick & Sherman, 2003) and riddles (Luo & Niki, (Sternberg & Davidson, 1995). This universally accepted conception of 2003a) are pure verbal problems. Anagrams require the rearrangement creativity could also include the latest definition of insight: “…areorga- of letters to create a new word, whereas riddles, in the context of insight nization of the elements of a person's mental representation of a stimu- research, are phrases with double or veiled meaning, in which one has lus, situation or event to yield a nonobvious or nondominant to guess the answer. interpretation” (Kounios & Beeman, 2014). Moreover, creativity and in- Brainteasers are verbal puzzles with practical content, which require sight seem to share many essential characteristics (Martindale, 1999)as unconventional thinking with given constraints in mind (Sheth, defocused attention, unconscious processing (Schooler & Melcher, Sandkühler, & Bhattacharya, 2009). 1997; Simonton, 2001) and less prefrontal activation (for details about The Chinese logogriphs are a particular type of riddle in which the insight correlations see the next paragraph). answer is a Chinese character that indicates a phrase, a Chinese proverb Thus, the Aha! moment seems to represent a specific sub-process in or a sentence in a poem that must to be solved by the addition, subtrac- which creative cognition could reaches consciousness (Aldous, 2007; tion or substitution of strokes, after having understood the riddle's im- Dietrich & Kanso, 2010). Supporting this theory, evidence of positive plicit meaning (Wang et al., 2009). correlation between performance at insight and creative tests (e.g. Binarized images are two-tone (usually black and white; Giovannelli CRA and Drawing Production: r = .274, p b .001; CRA and Alternative et al., 2010) pictures representing low-resolution version of an original Uses: r = .275, p b .001; Rebus Puzzle and Drawing Production: r = one. They appear meaningless at first sight, while becoming recogniz- .307, p b .001; Rebus Puzzle and Alternative Uses: r = .211, p b .05; able after a single exposure to its original version (i.e., priming). This Mourgues et al., 2014) as well as on other tests assessing different cog- purely visual task has been specifically used to assess the perceptual nitive abilities not related to analytical reasoning (e.g., perception) is learning component of Eureka moment. available (Bowden et al., 2005). In addition, insight is not involved in Mednick's RAT (Mednick, 1962) was created in the 1960s to mea- all phases of creative thinking (e.g. the critical evaluation of an idea), sure creative convergent thinking and is often used to assess creativity and is not a necessary feature of creative thinking, which could also in general (e.g. Gibson, Folley, & Park, 2009). It is also employed to esti- arise from an analytical process based on a defined multi-step approach mate insight problem-solving capacity (Cerruti & Schlaug, 2009; (Mumford & Whetzel, 1996). Furthermore, not all insight moments lead Razumnikova, 2007), since it was demonstrated that the performance to a creative process if we refer to the broader definition of insight, in- on RAT correlates with score at classical insight problem-solving (i.e. an- cluding perception and language comprehension. This could explain agrams correlation coefficient r = .55, Schooler & Melcher, 1997). The the very low correlations between insight and creativity tasks, RAT inspired the creation of the CRA by Bowden and colleagues, one supporting a more segregated view of these processes. Finally, as previ- of the most widely used tests to assess insight performance (Bowden ously mentioned, tasks used to assess creativity and insight are not et al., 2005). Both RAT and CRA stimuli consist of three words and the completely and clearly separable, thus such weak positive correlation subject has to find a fourth related word. However, while in the CRA could simply be the consequence of measurement error. Most creativity the solution must form a compound word with the others (i.e. “crab, tasks, such as the Alternate Uses Task, require divergent thinking, the pine, sauce”, solution: “apple = crabapple, pineapple, applesauce”), this ability to generate multiple solutions for an open-ended problem constraint is not present in the RAT (i.e. “falling, actor, dust”, solution: (Guilford, 1967). On the other hand, many insight tasks rely only on “star = falling star, movie star and stardust”). convergent thinking, the ability to find a single correct solution for a The advantages of RAT and CRA are pivotal for insight research: problem, or on a combination of the two (Abraham & Windmann, (i) they can be solved in just a few seconds by many people, (ii) 2007). However, CRA and RAT problem-solving could be divided into can be presented in a small visual space, (iii) the solution is a unique a first process involving divergent thinking, which is needed to explore word, facilitating scoring, (iv) there are many degrees of difficulty, the possible connections between the available stimuli, followed by a (v) several items, and, an important feature, (vi) they can be solved second step in which the subject converges to a single solution (conver- with both insight or analytical problem-solving processes (the strat- gent thinking). Therefore, tests of creative thinking often involve an ini- egy being used can be subjectively reported), thus allowing for a tial step requiring insight moments. In order to fully disentangle the proper comparison between these distinct mental mechanisms, in nature and organization of insight and creativity, a more in-depth com- order to reveal the unconscious restructuring characterizing the ini- prehension of their neurophysiological underpinnings should constitute tial stages of insight elaboration. This possibility marked a break an absolute priority (see also “The insight network” in the discussion from the old conception of insight tasks, for which the type of prob- paragraph). lem was the key determinant of the occurrence of an insight or an analytical mental operation (Bowden & Jung-Beeman, 1998; Bowden & 2.6. Multifactorial nature of Aha!: intelligence, memory, attention, mood, Jung-Beeman, 2003; Bowden & Jung-Beeman, 2007). According to cognitive control and sleep this innovation, CRA score correlates with two and three- dimensional classical insight tasks, (respectively r = .549 and r = As one of the most complicated and volatile cognitive phenomena, .430; Mourgues, Preiss, & Grigorenko, 2014). However, correlations insight has captured the attention of different fields of cognitive neuro- between classical and new insight tasks are not very strong, science and has been linked to diverse features of human behavior. For 104 G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 instance, the role of intelligence, attention levels, cognitive control, problems relying on insight strategies (Subramaniam, Kounios, mood, and sleep quality have been investigated, introducing numerous Parrish, & Jung-Beeman, 2009). In addition, mindfulness meditation, complex, but intriguing, scenarios. previously shown to have effects on mood modulation and introspec- tion levels (Santarnecchi et al., 2014), seems to correlate with perfor- 2.6.1. Intelligence and memory mance in solving rebus puzzles (Ostafin & Kassman, 2012). A correlation between insight and intelligence was demonstrated Associations between mood and insight abilities can also be reflected in children (Bermejo et al., 1996), with high-intelligence performers in event-related potentials (ERPs), such as the N400 component, classi- showing a stronger disposition to using insight problem-solving cally thought to reflect the semantic processing and integration of a than lower-intelligence performers. In adults, insight abilities were word in the provided context (McPherson & Holcomb, 1999). The am- found to strongly correlate with general fluid intelligence measures plitude of N400 is inversely related to the relationship of a provided (Sternberg & Davidson, 1995). Subsequent studies showed intelli- word in its context (Kounios, 1996): interestingly, positive mood repre- gence scores explaining almost 75% of variance in insight scores sents a good modulator of its amplitude (i.e. it induces a smaller N400; (Paulewicz et al., 2007), even though doubts about the construct va- Federmeier, Kirson, Moreno, & Kutas, 2001; for further details about lidity of the tasks used to assess the two abilities has been raised due N400 see the paragraph about ERPs). Therefore, positive mood may fa- to excessive overlap in task mechanics. In addition, a lesion study cilitate the accessibility of weak activation, making the provided stimuli (Barbey, Colom, & Grafman, 2013) revealed that psychometric intel- (words) more related to their semantic context. Another explanation is ligence strongly predicts cognitive flexibility, a key component of in- that positive mood modulates attention and cognitive control, facilitat- sight problem-solving (Subramaniam, 2008). Different aspects of ing perception and processing of external stimuli (Kounios & Beeman, WM were found to correlate with insight problem-solving, including 2014). spatial WM (Chein, Weisberg, Streeter, & Kwok, 2010), verbal short- term memory (Fleck, 2008), WM storage and processing (Murray & 2.6.4. Sleep and cognitive control Byrne, 2005), verbal spans and spatial storage capacity (Gilhooly & Finally, roles for sleep and cognitive control have been proposed Fioratou, 2009). Such a relationship is thought to be relevant in as well. Linked to many scientific discoveries (e.g., the periodic order to evaluate insight reasoning within the two dominant theo- table of Mendeleyev), sleep is likely to accelerate mental ries,i.e.theSpecial Process or the Business as Usual view restructuring of given information, a crucial component of insight (Chudersky, 2014;seethe“Defining processes” paragraph). Accord- problem-solving (Wagner, Gais, Haider, Verleger, & Born, 2004). As ingly to the former, WM should not correlate with insight (Van for cognitive control, Wieth and colleagues have recently shown Stockum & DeCaro, 2013;forareviewseeWiley & Jarosz, 2012), that participants perform better in tasks of insight problems during whereas the second view predicts a link between WM and insight their non-optimal mental times of day assessed by the Morningness tasks as shown for other high order cognitive functions such as crea- Eveningness Questionnaire (Wieth & Zacks, 2011). Intriguingly, this tivity and abstract reasoning (Chein et al., 2010; Murray & Byrne, result suggests that less cognitive control (i.e. less prefrontal inhibi- 2005). Following recent work by Chudersky (2014), factorial analy- tory dynamics in general) actually represents the optimal cognitive sis on a large dataset of insight, WM, and analytical reasoning tasks context to release constraints on weak semantic information and suggest that WM accounts for one third of variance in insight ease Eureka moments. problem-solving abilities. Considering both WM and analytical reasoning, two thirds of insight variance can be explained. In addition, 2.7. Overview of neurophysiological evidence about insight processes while both components of WM (storage capacity and executive control) seem to be involved in insight, reasoning and insight appear to In order to provide a comprehensive picture of the neural underpin- be two separable variables (i.e. they shared only 26% of variance). nings of insight, we first performed an original quantitative meta- Overall, both theories might fit the data, with a significant WM con- analysis of available fMRI data. We also examined existing EEG evidence tribution to insight problem-solving, but still a significant amount of involving both spontaneous brain oscillations and ERPs. Finally, a dis- unexplained variance even when reasoning is considered. cussion of the evidence from studies involving brain neuromodulatory approaches such as repetitive transcranial magnetic stimulation 2.6.2. Attention (rTMS) and transcranial electrical stimulation (tES) is offered. High attentional levels are commonly considered a typical feature of highly creative subjects (Carson, Peterson, & Higgins, 2003; 3. Methods Mendelsohn & Griswold, 1966; Rowe, Hirsh, & Anderson, 2007). In- terestingly, while high-insight subjects (participants using lots of in- 3.1. Literature search sight problem-solving strategies) seem to show high outward visual attention level in the resting state (Kounios et al., 2008,see We performed a literature search in PubMed and Google Scholar da- Discussion section and Fig. 5), a switch to more inward attention is tabases, without temporal restrictions, to retrieve potentially relevant observed during the mental preparation time (Kounios et al., articles. To specify the object of the present review, terms such as “func- 2006), which lasts until just before the intuition of a solution tional magnetic resonance imaging”, “electroencephalography”, “mag- (Jung-Beeman et al., 2004). This finding was recently confirmed netoencephalography”, “transcranial magnetic stimulation”, using eye-tracking (Salvi, Bricolo, Franconeri, Kounios, & Beeman, “transcranial direct current stimulation”, “transcranial alternating cur- 2015). Finally, a modulation of insight ability in relation to specific rent stimulation”, “transcranial random noise stimulation” and related attention patterns has been proposed (Wegbreit, Suzuki, abbreviations (fMRI, EEG, MEG, TMS, tDCS, tACS, tRNS) were individual- Grabowecky, Kounios, & Beeman, 2012), with subjects reaching ly combined with insight-related keywords such as “Insight”, “Insight higher scores in CRA problems after completing a broad attention- problem solving”, “Insight divergent thinking”, “Divergent thinking”, priming task. “Eureka”, “Aha”, “Aha reaction”, “Aha moment”. The searches for methods and research topic were combined with AND operator. Refer- 2.6.3. Mood ences of retrieved studies were examined for relevant publications as Mood has been found to influence insight processes as well: in line well. We intentionally excluded (i) studies including patients with or- with previous studies (Amabile, Barsade, Mueller, & Staw, 2005; ganic illness, (ii) exploring magic ideation, (iii) focusing solely on be- Estrada, Joung, & Isen, 1994; Isen, Daubman, & Nowicki, 1987; Rowe havioral data, (iv) review papers, (v) studies not mentioning insight in et al., 2007), participants with higher positive mood solved more CRA their abstract unless they used an insight task such as CRA, RAT or G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 105

Table 1 ERPs and fMRI studies. A complete list of all publications included in the fMRI and ERPs sections of the paper are displayed. Sample size, the speciﬁc task being used and a summary of the main ﬁndings are reported. Note: g. = gyrus; HFF condition = heuristic prototypes presented with highlighted functional features labeled above the prototype knowledge; OHFF condition = heuristic prototypes presented without highlighted functional features; SUSL = the base logogriph and target logogriph share some word; STSL = the base and target logogriphs do not have any words in common; BSL = the base and target logogriphs are all simple character-generation tasks; DLPFC = dorso-lateral prefrontal cortex; ACC = anterior cingulate cortex; PCC = posterior cingulate cortex; PFC = prefrontal cortex; RAT = Remote Associates Test; CRA = Compound Remote Associates problems (see Tables S1 and S2 for further details about experimental protocols).

Sample size Authors (mean Method Task Main results age, years)

fMRI studies Aziz-Zadeh et al., 1 12 (26) fMRI Anagrams Bilateral insula, Broca's area, right ventral prefrontal cortex, anterior cingulate cortex 2009 16 Technical problems with Left dorsolateral prefrontal gyrus, left angular gyrus (more significantly activated when presented 2 Dandan et al., 2013 fMRI (22.38) prototypes with related prototypes than with unrelated prototypes) Learning 17 Middle temporal gyrus, middle occipital gyrus (more activated under the HFF condition compared 3 Hao et al., 2013 fMRI prototypes-solving (22.1) with the OHFF condition) problems Jung-Beeman et al., 13 Increased activity in right hemisphere anterior superior temporal gyrus (RH aSTG), left medial 4 fMRI CRA 2004 (18–29) frontal gyrus, left posterior cingulate, bilateral amygdala or parahippocampal gyrus Bilateral superior frontal gyrus, bilateral middle frontal gyrus, left anterior cingulate cortex, right precentral gyrus, bilateral inferior frontal gyrus, right cingulate gyrus, bilateral superior temporal 7 5 Luo and Niki, 2003a fMRI Riddles gyrus, right hippocampus, right subgyral, left caudate, left inferior temporal gyrus, left inferior (20–22) parietal lobule, left superior parietal lobule, right precuneus, bilateral occipital lobe (activations shown for the presentation of the correct answer) Luo, Niki and 13 ACC, left lateral PFC, bilateral cingulate gyrus, left insula, inferior frontal gyrus, left middle frontal 6 fMRI Ambiguous sentences Phillips, 2004a (26.7) gyrus, left precuneus, left superior, frontal gyrus, left sub-gyral, left inferior temporal gyrus (i) Vs the resting baseline, solving of “cerebral gymnastics” puzzles (condition A): anterior cingulate gyrus, medial frontal gyrus, bilateral inferior frontal gyrus, right precentral gyrus, left postcentral gyrus, right superior, inferior temporal gyrus, left middle temporal gyrus, left inferior parietal lobule, Luo, Niki, and 21 7 fMRI Chinese puzzles left thalamus, left red nucleus, right medial globus pallidus, left angular gyrus, left precuneus; (ii) vs Phillips, 2004b (21-35) the resting baseline, solving of “homophone” puzzles (condition B): bilateral inferior frontal gyrus, bilateral insula, left anterior cingulate gyrus, left inferior parietal lobule, left supramarginal gyrus, left angular gyrus, left and right red nucleus, left precuneus (i) NSI (new scientific innovations) − OSI (old scientific innovation): lingual gyrus; OSI − NSI: right middle temporal gyrus, left medial frontal gyrus, left posterior cingulate gyrus, right thalamus, left 30 Heuristics prototypes 8 Luo et al., 2013 fMRI superior temporal gyrus; (ii) NSI − OSI: left precuneus, lingual gyrus; OSI − NSI: right medial frontal (23.5) and relative problems gyrus, right medial frontal gyrus, left inferior frontal gyrus, left middle temporal gyrus, left superior-parietal lobule, right supramarginal gyrus, right inferior parietal lobule 16 9 Qiu et al., 2010 fMRI Chinese logogriphs Precuneus, left inferior/middle frontal gyrus, inferior occipital gyrus, cerebellum (22.6) 16 Left middle/medial frontal gyrus, left middle/superior temporal gyrus, right cerebellum, bilateral 10 Tian et al., 2011 fMRI Chinese logogriphs (22.6) claustrum, left postcentral gyrus (i) Unconstrained vs. baseline trials: right ventral lateral PFC, occipital–parietal sulcus, right insula, left occipital gyrus, frontopolar cortex, bilateral superior parietal lobe, post central gyrus, Vartanian and Goel, 15, 11 fMRI Anagrams cerebellum; (ii) unconstrained versus semantically constrained trials: right ventral lateral PFC, right 2005 (26.5) superior parietal lobe, occipital–parietal sulcus frontopolar cortex, left superior frontal gyrus, right post central gyrus, cerebellum (i) Familiar (F) vs unfamiliar (U) chunks during tight chunk decomposition: ACC, bilateral inferior Wu, Knoblich, and 14 frontal gyrus; (ii) F-tight minus U-tight: activation in prefrontal areas; (iii) F-loose minus U-loose: 12 fMRI Chinese characters Luo, 2013 (19–25) right cingulate cortex, right middle frontal gyrus, right superior frontal gyrus, left superior frontal gyrus, left middle frontal gyrus, left inferior frontal gyrus, left fusiform gyrus (i) Insight (I) vs non-insight (NI) in the early period of solution: left superior temporal pole, inferior temporal gyrus, ACC, middle frontal gyrus, right inferior frontal gyrus, parahippocampal gyrus, bilateral gyrus, middle temporal gyrus; (ii) I vs NI in late period: left olfactory, middle frontal gyrus, 17 ACC, medial frontal gyrus, inferior parietal gyrus, right putamen, amygdala, bilateral middle 13 Zhao et al., 2013 fMRI Chinese riddles (23.6) temporal gyrus, hippocampus, angular gyrus; (iii) Timecourse activity: left and right middle temporal gyrus, left ACC, middle frontal gyrus (started at the beginning of the answer presentation and continued until the riddle's solution); left hippocampus and right amygdala (started just before the riddle's solution)

ERPs studies Large parietally focused SPW (slow positive wave), frontocentral P3a was larger, anterior N1 1 Lang et al., 2006 26 (24) ERPs Number reduction task amplitudes were larger, P300 complex decreased (solvers vs nonsolvers) (i) More negative N400 in centro-parietal scalp in STSL respect to SUSL and BSL; (ii) more positive 13 2 Luo et al., 2011 ERPs Chinese logogriphs P900–1700 ms in SUSL and STSL respect to BSL; (iii) more negative N1100–1300 over posterior scalp (22.4) regions in STSL respect to SUSL 14 3 Mai et al., 2004 ERPs Chinese riddles More negative N250–500 ms, maximal amplitude was at Cz (22.2) (i) Aha answer (AA)/Uncomprehended answer (UA) vs No-aha answer (NA): more negative ERP deﬂection (N250 and 400 ms); (ii) AA minus NA and UA minus NA: peak amplitude and latency 12 4 Qiu et al., 2006 ERPs Chinese logogriphs between 250 and 400 ms at Fz, Cz and Pz sites, maximal amplitude was at Pz, latency was 320 ms (21.4) (negative component N320 in central posterior region, mainly at right temporal parietal); (iii) difference wave AA − NA: dipole one located near the ACC, dipole 2 located near the thalamus 18 More positive P200–600, generator in left superior temporal gyrus and in parieto-temporo-occipital 5 Qiu et al., 2008 ERPs Chinese logogriphs (22.3) cortex; more negative N1500–2500 in left frontal regions; generator of N1500–2000: ACC, generator

(continued on next page) 106 G. Sprugnoli et al. / Intelligence 62 (2017) 99–118

Table 1 (continued)

Sample size Authors (mean Method Task Main results age, years)

of N2000–2500: PCC PWI (problems with impasses) elicited a greater anterior P2 than POI (problems without impasses), Shen, Liu, Yuan, 13 frontocentral P2 emerged earlier in peak latency and was larger in amplitude for PWI, difference 6 Zhang, and Luo, ERPs Chinese riddle problems (23–28) wave in the time course of 620–800 ms has seven neural generators, most important was left 2013 superior frontal gyrus and near the left cuneus Insight problems vs routine problems: more negative N300–800, more negative N1200–1500 over 12 fronto-central scalp regions, more positive P300–800 and more positive P1200–1500 over 7 Wang et al., 2009 ERPs Chinese logogriphs (21.4) parieto-occipital scalp regions; dipoles location: ACC and parahippocampal gyrus for 300 and 800 ms, in parahippocampal gyrus and the superior frontal gyrus for 1200 and 1500 ms 14 the amplitude of the P3a was significantly larger when familiar-tight chunks had to be decomposed 8 Wu et al., 2013 ERPs Chinese characters (19–25) than in all other conditions in centro parietal region (P350-650 ms) Xing, Zhang, and 12 Insight condition vs non-insight conditions: more negative N300–500 for most of the scalp regions, 9 ERPs Chinese logogriphs Zhang, 2012 (23.4) more positive P600–1100 in frontal, fronto-central and central scalp regions Insight solutions vs search solution: more positive P400–600, more positive LPC between 640 and 12 10 Zhang et al., 2011 ERPs Chinese characters 780 ms; in 400–600 ms interval: source on the fusiform gyrus; in 640–780 ms interval: source on (22.3) right superior temporal gyrus Chinese 13 BMS (breaking mental set condition) respect to REP (repetition condition): more positive P500–700 11 Zhao et al., 2011 ERPs character-generation (21.9) in centro parietal scalp region, more positive P900–1300 over centro-parietal scalp regions task (i) Semantic (SR) and homophonic punny riddles (HR) elicited a lower amplitude vs control 17, condition at center and at centro-parietal regions and a N350–500 over the central scalp; (ii) SR 12 Zhao et al., 2014a ERPs Chinese riddles (23.3) induced a positive deflection over temporal cortex at 400 to 0 ms before riddles were solved; (iii) HR induced a positive deflection over the temporal cortex and a negative one in left frontal cortex

riddles, and (vi) studies not reporting fMRI activations coordinates in of the inter-subject and inter-study variability usually observed in neuro- MNI or Talairach coordinates space (this criterion was used only for imaging experiments, rather than applying an a priori full-width half neuroimaging studies). The final selection comprised thirteen studies maximum (FWHM) kernel. Therefore, the number of participants in a related to fMRI, one study related to rTMS, twenty studies investigating given study influenced the spatial extent of the Gaussian function used. insight using EEG, four studies related to tDCS (Figs. S1 and S2). Tables 1, GingerALE first modeled the probability of activation over all studies at 2, S1, S2, S3, and S4 provide an overview of the papers included in the each spatial point in the brain, returning localized “activation likelihood review for each investigated methodology. estimates” or ALE values. As a second step, ALE values were compared to a null distribution created from simulated datasets with randomly 3.2. fMRI studies placed foci in order to identify significantly activated clusters (1000 per- mutations). Following Eickoff and colleagues' arguments supporting a For each fMRI study, the following information was retrieved: better balance between sensitivity and specificity for Cluster-based cor- (i) number of subjects, (ii) mean age, (iii) experimental design, (iv) cog- rections over False-Discovery-Rate (FDR) and Family Wise Error (FWE) nitive task details, and (v) main results (Tables 1 and S1). Data about the ones (Eickhoff et al., 2012), we applied a cluster correction for multiple specific activation foci was also collected and included in a quantitative comparisons with an uncorrected p b 0.001 threshold for cluster- activation likelihood estimation (ALE) analysis for the identification of formation and a p b 0.05 for cluster-level inference. Only clusters with a the most commonly reported brain regions involved in insight process- size exceeding the cluster size recommended by ALE were reported es. For details about the computation of the ALE meta-analysis maps, see (range in our sample = 250–1000 mm3). Anatomical labels of final clus- below. Results of the ALE map computation are reported in Fig. 2. ter locations were provided by the Talairach Daemon (http://www. talairach.org/daemon.html) and available as part of the GingerALE output. 3.2.1. Computation of the ALE meta-analysis maps Entering the insight meta-analysis were thirteen studies with a total of The quantitative evaluation of spatial fMRI patterns published about 236 Talairach/MNI coordinates (Table 1). ALE maps were computed for insight was carried out using the ALE technique implemented in the each of the 13 studies, and areas of significant activity along with their GingerALE software v2.3.2 (www.brainmap.org; Turkeltaub et al., Talairach coordinates are reported in Table 3. Given the relatively low 2012; Eickhoff et al., 2009a; Eickhoff, Bzdok, Laird, Kurth, & Fox, number of studies, the results are shown for the average activation map 2012). This method yields a statistical map that indicates the set of built by averaging activation patterns for different insight tasks (Fig. 2). brain voxels that are more active than would be expected by chance. Each ALE map was visualized using MriCronGL 64 (http://www. Different from within-study statistical parametric mapping analysis, mccauslandcenter.sc.edu/CRNL/tools/). where every voxel in the image space is tested against a null hypothesis of no activation, the ALE method assumes that for each study of interest 3.3. EEG studies there is a given spatial distribution of activity and an associated set of maximal coordinates. Therefore, the algorithm tests to what extent For each EEG study, we retrieved: (i) number of subjects, (ii) mean the spatial locations of the activation foci correlate across independently age, (iii) experimental design, (iv) cognitive task used and (v) main re- conducted fMRI studies investigating the same construct. sults regarding brain oscillations (changes in topography and power/co- Coordinates collected from studies that were reported in Talairach herence) and ERPs (changes in power; Tables 1, 2, S2 and S3). spacewereconvertedintoMNIspaceusingthe“tal2mni” algorithm im- Moreover, studies showing correlations between spontaneous brain func- plemented in GingerALE. First, activation foci from each study were tioning and insight abilities were analyzed as well. Given the impossibility modeled as Gaussian distributions and merged into a single 3D volume. to run a quantitative analysis of published EEG results about spectral The ALE algorithm modeled the spatial uncertainty (Eickhoff et al., power or coherence, EEG evidence has been collected, discussed and 2009a; Eickhoff et al., 2012) of each activation focus, using an estimation summarizedbytheauthorsasshowninFigs.3,4and5. G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 107

Table 2 NiBS and brain oscillations studies. A complete list of all publications included in the NiBS and brain oscillations sections of the paper are displayed. Sample size, the specific task being used and a summary of the main findings are reported (see Tables S3 and S4 for further protocols' details). Note: Δ = delta activity; α = alpha activity; θ = theta activity; β = beta activity; γ = gamma activity; ERS = event-related synchronization for higher task-related power compared to the baseline (see Figs. 4 and 5 for the specific oscillatory frequencies); DLPFC = dorsolateral prefrontal cortex; ATL = anterior superior temporal lobe; tDCS = transcranial Direct Current Stimulation; rTMS = repetitive Transcranial Magnetic Stimulation; RAT = Remote Associates Test; CRA = Compound Remote Associates problems.

Sample size Authors (mean Method Task Main results age, years)

Brain oscillations studies Higher EEG spectral power of the Δ and θ bands in the right anterotemporal area; higher power of Δ band in the left anterotemporal area; coherence in Δ, θ, α1, α2, and β1 frequency bands lower in 30 1 Danko et al., 2003 EEG RAT (variation) right prefrontal–left anterotemporal electrodes; in the Δ, θ, α1, and α2 bands coherence lower in students left frontal–right anterotemporal areas; for Δ, θ, and α1 bands lower in left prefrontal–right anterotemporal lobes Jung-Beeman et al., 19 Sudden burst of γ band in anterior right temporal electrodes, α band activity over right posterior 2 EEG CRA 2004 (18–29) parietal cortex (i) Sustained power decrease in α and β bands after stimulus onset in all 3 conditions (NN: no-recognition + no recognition; NR: no-recognition + recognition; RR: recognition + 13 recognition); (ii) larger β power decreases in NR minus NN for centro temporal cluster (source in 3 Minami et al., 2014 EEG Binarized images (24.6) temporal, parietal, subcortical regions) and in the NR − RR for parietal cluster, (sources in parietal regions and in subcortical regions); (iii) γ band: signiﬁcant cluster in frontal regions in NR vs RR; (iv) NR vs NN and NR vs RR overlapped in medial parietal cortex, mainly precuneus Widespread enhancement of power and coherence in the β2, the θ1 power increase in frontal 39 cortex; increased desynchronization of the α1, α2 mainly over posterior cortex; α1 coherence 4 Razumnikova, 2007 EEG RAT (17–20) decrease in prefrontal sites; originality scores of the verbal associate: increase of α1 coherence focused in fronto-parietal regions of both hemispheres in the β2 and in left parieto-temporal loci Sandkühler and 21 Strong γ band responses at parieto-occipital regions and later θ frequency band cluster; strong 5 EEG CRA Bhattacharya, 2008 (26.4) upper α ERS in right temporal regions; decreased α power in right prefrontal cortex Reduction in β power over the parieto-occipital and centro-temporal regions on all four conditions: correct vs. incorrect solutions, solutions without vs. with external hint, successful vs. unsuccessful 18 6 Sheth et al., 2009 EEG Brainteasers utilization of the external hint, and self-reported high vs. low insight. “Aha!” versus “non-Aha!”: (21.2) reduction in power in β frequency band in central-parietal regions, increased power in lower α frequency band (difference was signiﬁcant over the right frontal region in γ frequency band) (i) Insight preparation: greater neural activity (less α) over mid-frontal, left temporal, right Mental 7 Kounios et al., 2006 19 CRA temporal, right inferior frontal, and bilateral somatosensory cortex; (ii) noninsight preparation: preparation-EEG less α power than insight preparation in occipital cortex Low Insight (LI) group had more high α than the High Insight (HI) group over occipital cortex. HI showed: (i) HI subjects showed less β1 EEG power over occipital midline (EC); (ii) EO: less occipital β1 power with a broader distribution; (iii) greater low α power at left inferior–frontal and anterior–temporal sites, less power in right dorsal–frontal region; (iv) greater β2 power at right inferior–frontal and anterior–temporal sites, less β2 power in left occipital and parietal sites; 8 Kounios et al., 2008 26 (22) Resting-state-EEG Anagrams (v) EC: greater β3 power at right frontal–temporal sites and less power at left parietal sites, EO: greater β3 power at right parietal and temporal–occipital electrodes; (vi) EC: greater γ power at right inferior–frontal and left temporal electrodes and less power for HI subjects in right inferior–parietal sites; EO: more γ band power at right-parietal electrodes and a weaker left-parietal area

NiBS studies (i):18 Cerruti and (25.5) (i) Increase in performance after anodal tDCS on left DLPFC (F3) vs sham and cathodal stimulation; 1 tDCS RAT Schlaug, 2009 (ii): 12 (ii) F3 anodal stimulation produced a significant effect on RAT; right DLPFC stimulation did not (25.4) Matchstick Chi and Snyder, (i) 60% of participants solved matchstick with anodal tDCS (right ATL); (ii) 20% of participants did 2 60 (22) tDCS arithmetic 2011 so with sham stimulation; (iii) opposite polarities did not facilitate performance problem Chi and Snyder, 22 Nine-dot (i) 0/11 participants with sham stimulation solved the task; (ii) 5/11 participants solved it with 3 tDCS 2012 (19–63) problem anodal tDCS on right ATL; (iii) opposite polarities of stimulation did not affect performance 21 Anodal tDCS (left DLPFC) enhanced solution recognition for difficult trials, larger effect for 4 Metuki et al., 2012 tDCS CRA (23.1) participants with lower approach motivation (i) 10 Hz rTMS delivered at the same time of undegraded images' presentation on right and left Giovannelli et al., 33 5 rTMS Binarized images intraparietal sulcus reduced the percentage of binarized images correctly identified after learning 2010 (24.2) phase (30′); (ii) no effect with rTMS delivered 2 s after undegraded images' presentation

3.4. Non-invasive brain stimulation studies search and a summary of main findings regarding NiBS are reported in Fig. 6. For each tES study, we retrieved: (i) number of subjects, (ii) mean age, (iii) experimental design, (iv) tES montage (target and reference 4. Results electrodes’ positions, online/offline stimulation), (v) type of cognitive task being modulated, and (vi) main results (Tables 2 and S4). For 4.1. Insight-related fMRI activations rTMS literature, we extracted: (i) number of subjects, (ii) mean age, (iii) experimental design, (iv) TMS setup (Sham method, stimulation The results highlight a complex network composed of the anterior site), (v) task specifics, and (vi) main results. Results of the literature cingulate cortex (ACC), prefrontal and parietal lobes, claustrum, 108 G. Sprugnoli et al. / Intelligence 62 (2017) 99–118

Fig. 2. Insight-related fMRI activations. The results of cluster-based statistics performed on the studies retrieved using the keywords specified in the supplementary methods section of the paper are displayed. The map refers to studies assessing insight problem-solving without any discrimination for trial or task types. Shown in both axial (A) and coronal 3D (B) views, it is therefore a non-specific, global representation of insight-related cognitive processing in the brain. The quantitative meta-analysis highlighted a network of brain regions consisting in: 1 = left premotor/supplementary motor area, 2 = left middle temporal gyrus and precuneus, 3 = right superior frontal gyrus and left cingulate gyrus, 4 = left claustrum, 5 = left middle temporal and occipital gyri, 6 = uvula (cerebellum), 7 = left precentral gyrus/frontal eye fields, 8 = right insula, 9 = left insula, 10 = right precuneus, 11 = right middle temporal gyrus, (uncorrected p b 0.001 threshold for cluster-formation, corrected p b 0.05 for cluster-level inference, threshold for cluster size = 250 mm3). Given the functional roles commonly reported for these regions, mostly involved in executive functions and abstract reasoning processes, a qualitative analysis of the overlap with resting-state fMRI networks has been completed (C), resulting in major similarities between the “insight network” and the anterior salience and the left executive functions networks. A complete set of coordinates for each cluster is available in Table 3. Both weighted binary and cluster-based versions of the meta-analysis map are available for download as nifti “.nii” volumetric files at http://www.tmslab.org/santalab.php. temporo-occipital regions, middle temporal gyrus and insula (Fig. 2A& parcellation of the insight network in 11 functional nodes that can be B, Table 3). Surprisingly, the majority of activations are localized in left used as separate ROIs for volumetric and functional connectivity analy- hemisphere. More in detail they are: precentral gyrus, middle temporal sis, and an excel spreadsheet reporting coordinates for each cluster as gyrus, precuneus, cingulate gyrus, claustrum, middle occipital gyrus, well as size and anatomical labels. uvula (inferior vermis - cerebellum) and insula. As right activations, the map shows: superior frontal gyrus, insula, precuneus and middle 4.2. Insight-related ERPs temporal gyrus. Many insight regions are part also of other networks, as left executive functions network and creativity (see Fig. 2 and ERPs findings show a bilateral network, ranging from frontal lobe to Discussion section). The ALE meta-analysis map is available for down- the occipital pole (see Fig. 3). Most studies report widespread increased load as a nifti “.nii” volumetric file at http://www.tmslab.org/ESantarn. amplitudes of early positive and negative peaks resembling the P300 Files include a weighted map of the entire network, cluster-based and N400 component (latency ranges 200–800 and 250–800

Table 3 Results of ALE Meta-analysis. The results of cluster-based statistics performed on studies retrieved using the keywords speciﬁed in the supplementary methods section of the paper. Vol- ume; coordinates and local maxima for each cluster; as well as labels based on the Brodmann atlas and anatomical localization at cortical and subcortical level (lobe/gyrus/region) are provided (uncorrected p b 0.001 threshold for cluster-formation; corrected p b 0.05 for cluster-level inference; threshold for cluster size = 250 mm3).

Extrema value and Cluster Volume Weighted center Brodmann localization Hemisphere Lobe Gyrus/region number (mm3) area xyz xyz

2112 −44.58 3.96 29.38 0.0202 −42 2 28 BA 6 Left Frontal Premotor/supplementary motor area 1 0.0190 −50 6 28 BA 6 Left Frontal Premotor/supplementary motor area 1728 −28.52 −65.42 31.45 0.0193 −30 −62 32 BA 39 Left Temporal Middle temporal gyrus 2 0.0175 −26 −70 30 BA 31 Left Occipital Precuneus 1608 3.59 14.64 45.72 0.0195 6 14 50 BA 6 Right Frontal Superior frontal gyrus 3 0.0145 0 12 40 BA 32 Left Limbic Cingulate gyrus 4 1384 −33.41 17.68 −2.27 0.0281 −34 18 −2 Left Sublobar Claustrum 976 −49.05 −58.3 −3.19 0.0189 −50 −56 −2 BA 37 Left Temporal Middle temporal gyrus 5 0.0155 −48 −66 −6 BA 37 Left Occipital Middle occipital gyrus 6 440 −5.33 −79.75 −32.55 0.0165 −6 −80 −32 Left Cerebellum Uvula 7 440 −27.08 −0.93 55.99 0.0155 −26 0 56 BA 6/8 Left Frontal Precentral gyrus/frontal eye ﬁelds 368 39.29 7.35 13.77 0.0124 38 8 14 BA 13 Right Subcortical Insula 8 0.0111 44 2 16 BA 13 Right Subcortical Insula 9 328 −39.29 12.46 10.63 0.0146 −38 12 10 BA 13 Left Subcortical Insula 10 328 26.97 −69.03 47.44 0.0147 26 −70 48 BA 7 Right Parietal Precuneus 11 312 52.13 −56.13 −9.21 0.0141 52 −56 −10 BA 37 Right Temporal Middle temporal gyrus G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 109

Fig. 3. Insight-related ERPs. (A) Results of the review of ERPs detected during insight problem-solving (mostly using Chinese tasks, for details see Table 1 and S2), using color-coding for activity localized on the right or left hemisphere for each lobe, or both. The activity in each time window is expressed as follow: an “increase” in a specific ERP refers to a bigger response going in the direction of the ERP, which means an increased N400 (↑ N400) represents a more negative N400, while a decreased P300 (↓ P300) means a less positive P300. (B) Anatomical mapping of electric sources identified in various studies are reported (i.e. using LORETA or similar approaches), with yellow markers also highlighting sources roughly overlapping with regions identified in the ALE fMRI map shown in. Note: LORETA = Low Resolution Tomography Analysis.

respectively). Later positive and negative components are displayed, as Starchenko, & Bechtereva, 2003; Razumnikova, 2007). These experi- N1200–1500, in frontal, parietal and occipital lobes and P1200–1500 in ments used very similar tasks to assess insight (i.e. RAT and CRA) so parietal and occipital regions. the concordance of their results is not surprising; nevertheless, the experimental protocols were not identical. Danko created a new version 4.3. Brain oscillations during insight of RAT in which subjects had to ﬁnd a link between 12 words from different semantic ﬁelds; in Sandkühler et al., participants were given 45 s During insight problem-solving, a consistent decrease of α power in to solve CRA problems and an additional clue was provided if they were the right hemisphere and of coherence (bilaterally) is observed in the not able to solve the trials. Finally, Razumnikova's protocol followed the frontal lobe (Fig. 4; Sandkühler & Bhattacharya, 2008; Danko, canonical use of RAT problems, with participants solving each item with

Fig. 4. Brain Oscillation during Insight. Results of the review of brain oscillatory activity recorded during insight problem-solving. Color-code indicates hemispheric distribution. Details about the speciﬁc publications evaluated are also reported. Note: Δ = delta activity between 1.5 and 3.5 Hz; α1 = alpha activity between 7.5 and 9.5 Hz in (1) and between 8 and 10 Hz in (4); α2 = alpha activity between 10 and 12.5 Hz in (1) and between 10 and 13 Hz in (4); θ1 = theta activity between 4 and 6 Hz in (4); β1 = beta activity between 13 and 18 Hz in (1) and between 13 and 20 Hz in (4); β2 = beta activity between 18.5 and 30 Hz in (1) and between 20 and 30 Hz in (4); γ = gamma activity above 35 Hz up to 70 Hz; ERS: event-related synchronization for task-related power compared to the baseline (5). 110 G. Sprugnoli et al. / Intelligence 62 (2017) 99–118

Fig. 5. Brain Oscillations during Mental Preparation and Resting-State Activity. Brain oscillatory activity detected right before an insight solution or during spontaneous resting-state is displayed. Color-code indicates hemispheric distribution for each lobe. Details of the speciﬁc publications being evaluated are also reported. Note: EO = eyes open, EC = eyes-closed; α =alpha;θ = theta; β1 = beta activity between 13.00 and 17.75 Hz; β2 = beta activity between 18.00 and 24.75 Hz; β3 = beta activity between 25.00 and 29.75 Hz; γ =gamma.

no time limit (for further details see Tables 2 and S3). We must note that The α insight pattern is usually accompanied by an opposing pattern α activity seems to be important also for creativity, however in the op- in the β range (i.e. bilateral increase of coherence, Razumnikova, 2007) posite way: transcranial alternating current stimulation (tACS) in the α as well as by a bilateral increase in θ power (Razumnikova, 2007). Con- band (i.e. 10 Hz) delivered over bilateral prefrontal lobes during the versely, results about γ activity in the frontal lobe are characterized by Torrance Creativity Test improves creativity score, suggesting a beneﬁ- an increase of γ power restricted to right hemisphere (Sheth et al., cial role for α oscillations (Lustenberger, Boyle, Foulser, Mellin, & 2009) but a bilateral decrease is also reported in a different study Frohlich, 2015). While suggests a differential contribution of α activity (Minami, Noritake, & Nakauchi, 2014). However, we had to highlight for insight and creativity, it should be noted that the tasks used are that the protocol and the stimuli used in these two studies are very dif- very different. ferent: for example, Minami et al. analyzed the “perceptual insight”

Fig. 6. Non-Invasive Brain Stimulation and Insight. Results of the qualitative analysis of tDCS and rTMS (rTMS-10 Hz: excitatory effect) evidence when stimulation was applied during insight problem solving. Color-code indicates polarity of tDCS arranged by site of application. “Sham” stimulation refers to “placebo” stimulation. G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 111

Fig. 7. Correlates of the Insightful Mind. The most widely reported correlates of Aha! Moments are summarized, with different sets of results for experimental ﬁndings related to (A) the resting-state brain activity and the mental preparation phase preceding insight-related answers, as well as (B) evoked activity during actual Eureka moments. Results are color-coded accordingly to electrophysiological correlates related to brain oscillations, event related potentials and the overlap between EEG and fMRI patterns highlighted at the ALE meta-analysis.

using the binarized images, on contrary Sheth et al. administered brain- anagrams (Kounios et al., 2006; Kounios et al., 2008; Fig. 5). The study teasers to participants (for further protocol's details see Table S3). Over- on mental preparation (Kounios et al., 2006) focused on the stream of all, these activation patterns suggest a bilateral pattern with a right- brain activity occurring while the subjects elaborated about available in- lateralization only somewhat clearly present for α power. formation prior to an actual insight event (pre-Aha!). They reported a Regarding the parietal lobe, an increase in power for almost all the decrease of α power in frontal, parietal and temporal areas. These re- frequency bands (i.e. α, θ and γ; Jung-Beeman et al., 2004; Sheth gions, traditionally associated with cognitive control and semantic pro- et al., 2009; Sandkühler & Bhattacharya, 2008), and in coherence for cessing, suggest that such oscillatory patterns could be a sign of α1 (7, 5–9, 5 Hz) and β2(18–24.75 Hz) bands, is reported increased activity in regions related to semantic activation (Mashal (Razumnikova, 2007). As observed for the frontal lobe, a bilateral pat- et al., 2005; Mason & Just, 2004). Additionally, the occipital cortex pre- tern seems to define insight processes, with an exception for a right- sents an increase of α power, which could indicate a removal of focused lateralized increase in α power as the only evidence supporting the attention on external stimuli and an increase in more introspective idea of right-hemispheric dominance (the increase in α coherence is in- mental processes that aid the “insightful subject” in finding the most deed restricted to the left hemisphere; Razumnikova, 2007; Fig. 4). distant association between the stimuli at hand. The temporal lobe shows less consistent results, with studies dem- Intriguingly, oscillatory activities observed during mental prepara- onstrating different oscillatory activity including a burst of γ activity tion are in contrast with the resting-state results shown by Kounios (Jung-Beeman et al., 2004)andanincreaseofθ power in the right hemi- et al. (2008). Here, better insight skills seem to correlate with an in- sphere (Danko et al., 2003), as well as a bilateral decrease in Δ crease of α power in left frontal and bilateral temporal lobes, as well 1.5–3.5 Hz), θ, α and β coherence (Danko et al., 2003). Overall, these ac- as with its decrease in the bilateral occipital cortex (Fig. 5, Tables 2, tivations seem to be more lateralized, with a dominance of the right and S3). As for the rest of the frequency spectrum, a lateralized pattern hemisphere, especially in the γ band (Fig. 4). appears in the frontal lobe, with the enhancement of γ and β power re- Finally, occipital lobe activations are completely bilateral but studies vealed only in the right hemisphere. With all the limitations of an obser- are not consistent across the results provided. One study reports an in- vation based on two samples, it seems intuitive that the oscillatory crease in θ and γ power (Sandkühler & Bhattacharya, 2008), while pattern found during mental preparation is very different (almost the others report α desynchronization and a decrease in β power (Sheth opposite) from the resting-state activity of insightful subjects. et al., 2009; Razumnikova, 2007; Fig. 4). Nonetheless, different protocols and tasks could be considered responsible of these varying results. In 4.5. NiBS and insight fact, only two studies (Jung-Beeman et al., 2004Razumnikova, 2007) created a simple protocol in which subjects had to solve insight prob- Only five papers examined the application of TMS or tDCS, with var- lems without hints, through using RAT and CRA, with and without iability in the type of insight task and stimulation sites (Fig. 6, Tables 2 time limits, respectively (Table S3). and S4). The first study used tDCS over the prefrontal cortex to enhance insight problem-solving (Cerruti & Schlaug, 2009). In this experiment, 4.4. Brain oscillations during mental preparation and resting-state activity the authors demonstrated that anodal stimulation over the left dorsolateral prefrontal cortex (DLPFC) increased RAT scores in contrast to We found only two studies on mental preparation and resting-state cathodal or sham tDCS over the same site, which showed no effects on activity, both analyzing the performance in two verbal tasks: CRA and performance (anodal is supposed to induce an excitatory effect while 112 G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 cathodal leads to an inhibitory one, sham refers to placebo stimulation). potential interventions aimed at enhancing such a capability by means In a second experiment, the authors tested the effect of anodal tDCS on of NiBS. A discussion of the main findings highlighted during the review the right DLPFC, confirming the improvement in test scores after stimu- process will follow (see Fig. 7 for a summary), pointing out those con- lation of left prefrontal cortex (in both experiments the return electrode sidered as the most relevant features of insight in humans. was placed over the contralateral supraorbital region; Cerruti & Schlaug, 2009). In another study by Metuki and colleagues (Metuki, Sela, & 5.1. The insight network Lavidor, 2012), anodal tDCS over the left prefrontal cortex enhanced CRA performance compared to sham stimulation (with cathode over Examining the regions active during insight processes reveals a de- the right orbitofrontal cortex). However, this study did not follow the gree of overlap with other brain regions underlying divergent (e.g. cre- standard CRA procedure. In fact, participants were asked whether a pre- ativity) and convergent thinking abilities (e.g. fluid intelligence, sented word was the correct solution for the trial at hand. This might executive functions, Fig. 2). Firstly, activity in the middle frontal gyrus constitute a significant deviation from the canonical assessment of in- (MFG), inferior frontal gyrus (IFG) and ACC has been repeatedly associ- sight through CRA problems, therefore these results must be considered ated with memory, attention, inhibition, switching, language (Benn carefully. et al., 2014; Colom, Jung, & Haier, 2007; Roth, Serences, & Courtney, As for those papers showing effects during tES over the temporal 2006; Yin et al., 2012), general cognitive control (Miller & Cohen, lobe, the same montage was applied with anodal and cathodal stimula- 2001) and creativity (Boccia, Piccardi, Palermo, Nori, & Palmiero, tion respectively targeting the right and left anterior temporal lobe 2015). Furthermore, the dorso-lateral prefrontal cortex (DLPFC) has (Fig. 6; Chi & Snyder, 2012; Chi & Snyder, 2011). Even though insight been linked to problem-solving tasks exploiting analogous relationships was indexed by means of two different tasks (i.e. nine-dot and match- (Boroojerdi et al., 2001; Wharton et al., 2000) and relational complexity stick problems), both studies reported an increase in performance dur- (Kroger et al., 2002). The same regions were found activated also during ing anodal stimulation over the right hemisphere compared to sham creativity tasks (Boccia et al., 2015), suggesting a common substrate for stimulation or to the reverse electrical pattern (anode on the left tempo- executive functions in either creative or insight problem-solving pro- ral lobe, cathode on the right temporal lobe). Importantly, the nine-dot cesses (i.e. top-down control of attention and cognition; Beaty, problem is a single trial task with consequently important limitations in Benedek, Kaufman, & Silvia, 2015), rather than representing a specific the terms of results reliability: while none of participants solved the network for these two processes. nine-dot problem during sham stimulation, 40% of them correctly The same applies to temporo-occipital regions and to the middle solved the task within the 3 min of tDCS or immediately after (Chi & temporal gyrus, often reported in studies investigating embodied per- Snyder, 2012). While these results are promising, single trial tasks spective taking (Wang, Callaghan, Gooding-Williams, McAllister, & should be avoided during the assessment of complex cognitive func- Kessler, 2016), language and memory (Bogels, Barr, Garrod, & Kessler, tions such as insight, in favor of more reliable and reproducible tools, 2015), mental imagery (Kosslyn & Thompson, 2003), as well as creativ- e.g. the RAT or CRA. ity (Boccia et al., 2015), all functions in which long-term memory might Along the same line, the authors presented another study involving play a relevant role in restructuring available information (Fink et al., matchstick problems where only the 20% of participants solved the 2012). As for the overlap with creativity, the right fusiform gyrus was hardest trials (i.e. type 2) during sham tDCS, while the percentage of found selectively activated during musical improvisation tasks (Boccia success reached 60% during tDCS stimulation (Chi & Snyder, 2011). et al., 2015), in line with the evidence of its role in non-verbal test of as- The same pattern was reported for easier trials (type 3), where the per- sociative semantic knowledge (Mion et al., 2010). Additionally, the left centage increased from 45% during sham tDCS to 85% during anodal supplementary motor area/frontal eye fields seem to represent a com- tDCS over the right temporal lobe. mon area for insight and creativity networks. This is not surprising, con- Finally, we retrieved one study about rTMS and insight, evaluating sidering its anatomical connection of the DLPFC and visual cortices, in the perceptual recognition of degraded images as a measure of the frame of its possible role in the visually guided behavior (Wright & insight-perceptual ability (Fig. 6, Tables 2 and S4). Repetitive TMS at Lawrence, 2008). In addition, there is evidence for its implication in 10 Hz (known to induce an excitatory effect, Tang et al., 2015) was ap- the control of spatial attention (Moore & Fallah, 2004), top-down con- plied over the vertex as well as over the right and left lateral parietal trol of visual areas, as well as of many aspects of visual cognition cortices in three separate sessions. Results showed that right and left (Vernet, Quentin, Chanes, Mitsumasu, & Valero-Cabre, 2014), all essen- intraparietal sulcus stimulation delivered during stimulus presentation tial operations required for the internal search of a creative solution or reduced the percentage of binarized images (i.e. two-tone picture, insight moment. black and white) correctly recognized 30 min after the learning phase. As reported in Fig. 2, many areas are shared with the salience net- Even though it was a single observation, this data does not support work, such as the middle temporal gyrus, the claustrum, and the the idea of a dominant role for left or right parietal cortex in Eureka- precentral gyrus. The salience network is important for reallocating at- related processes. In line with the EEG findings, the nature of the stimuli tentional resources in relation to external inputs and internal brain supports both parietal lobes as a crucial hub of the perceptual-insight events (Bressler & Menon, 2010) and it seems involved in dynamic processing, especially during the consolidation phase, given that rTMS switching between the default mode network and the executive control showed no effect on recognition of degraded images when applied 2 s network (Sridharan, Levitin, & Menon, 2008). Considering the involve- after stimulus presentation. ment of all these networks in creativity and insight moments, salience regions could be fundamental to permit a dynamic switching between 5. Discussion opposite networks during these cognitive processes (Beaty et al., 2015). Brain imaging and electrophysiological studies have suggested the A systematic review of the available fMRI, EEG and NiBS literature of insula as a key node for interoception, as well as bodily and emotional insight has been provided, in an effort to integrate what is currently awareness (Craig, 2009), due to its extensive visceral-sensory inputs available about the neurophysiological basis of the Eureka moment. A from the periphery and reciprocal connections with limbic, somatosen- complex yet fascinating picture arises, with several models and theories sory, prefrontal and temporal cortices (Augustine, 1996; Mesulam & gaining support from experimental evidence, while other facets of in- Mufson, 1982). The insula has been specifically linked to the monitoring sight still remain obscure. Given the relevance of the insight process in of visceral functions of the body, and its possible role in interoception everyday life as well as in technological achievements and human prog- offers a possible basis for its involvement in all subjective feelings ress in general, it is of absolute importance to deepen our knowledge of (Nieuwenhuys, 2012), with recent evidence also showing structural this pivotal feature of human cognition. This is a pre-requisite to inform changes after interventions aimed at increasing self-awareness G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 113

(Santarnecchi et al., 2014). It is possible that insight (and creativity, see 5.2. The timing of an insightful solution: event-related potentials during Boccia et al., 2015) processing implies a balancing of externally and in- Eureka! ternally driven brain states, making the insula a key node for the modulation (i.e. suppression) of bodily sensation in favor of a more As shown in Fig. 3, results are apparently cloudy, showing several thought-centered monitoring. On the other hand, insula was linked to short and long-latency components with positive and negative polari- many stages of language comprehension/production (Ackermann & ties. However, the qualitative analysis suggests a major role for ERP ac- Riecker, 2004; Blank, Scott, Murphy, Warburton, & Wise, 2002; tivity involving midline regions in the frontal and parietal lobes, with no Eickhoff, Heim, Zilles, & Amunts, 2009b; Oh, Duerden, & Pang, 2014), standing left or right hemispheric dominance (Fig. 3, Tables 1 and S2). as articulatory planning and phonological recognition (Ardila, Bernal, Remarkably, however, ERP data is mostly based on brain activity during & Rosselli, 2014). In particular, anagrams were found to elicit a bilateral the Chinese logogriphs task (Table 1), which could limit the generaliza- (Aziz-Zadeh et al., 2009) and right lateralized (Vartanian & Goel, 2005) tion of these results to other insight processes. Here we first discuss the insula activation. Curiously, analytical solutions elicited only a left acti- negative components, which are thought to represent (i) cognitive im- vation (Aziz-Zadeh et al., 2009), supporting the bilateral involvement passe, (ii) the feeling of warmth and creations of new associations, of brain's network necessary for insight problem-solving (as theorized followed by positive components, which are thought to underlie: by Bowden and Beeman), in which left and right insula cooperate for (i) breaking of mental set and (ii) forming novel associations. the linguistic component of solving anagrams. In Vartanian and Goel (2005), the unconstrained trials elicit activation of right insula com- 5.2.1. Negative potentials around 400 ms pared to baseline trials in which the solution is presented to the partic- The most consistent result is a positive correlation between insight ipants. In addition, Adank (2012) finds insula activations during problem-solving and an N400-like peak recorded at frontal sites distorted speech comprehension, which could aid explaining Luo, Niki, (Fig. 3), with only one study reporting a negative correlation (Zhao, Li, and Phillips (2004a) insula activation in response to ambiguous Shang, Zhou, & Han, 2014a). Well-known in the context of semantic in- sentences. formation processing, the N400 typically has its maximum amplitude in The claustrum might constitute an underrated node of the insight the centroparietal lobe during the processing of written words and is network. This subcortical gray matter region, located lateral to the puta- interpreted as an ERP index of semantic processing and integration of men and medial to the insular cortex, is part of the basal ganglia circuit- a word in the current context (McPherson & Holcomb, 1999). An in- ry (Schmitt, Eipert, Kettlitz, Lessmann, & Wree, 2016). The claustrum's crease in the amplitude of N400-like components in insight could there- activity has been related to attention (deBettencourt, Cohen, Lee, fore reflect a greater effort of the brain in the integration of the stimuli Norman, & Turk-Browne, 2015; Fall, Querne, Le Moing, & Berquin, within a given context (Lau, Phillips, & Poeppel, 2008). Other interpre- 2015; Goll, Atlan, & Citri, 2015) and more specifically to interhemi- tations of this negative potential have been suggested. Mai, Luo, Wu, spheric communication between attention-related regions (Smith & and Luo (2004) proposed that this negativity (N380 with central Alloway, 2014). Little is known about the claustrum's specificfunctions, focus) might be a marker of breaking the mental set. This interpretation with some clinical evidence suggesting a possible role in maintaining stems from a dipole analysis (Mai et al., 2004), which localized the gen- consciousness. A recent study documented a case of “on-off” switching erator of the N380 wave in the ACC, an area that has been related to of level of consciousness resulting from stimulation of the claustrum in N200 and error-related negativity (ERN) components. Although Qiu an epileptic patient (Koubeissi, Bartolomei, Beltagy, & Picard, 2014). In et al. (2006) found a similar result (Aha-answers elicited a larger the context of insight-related processing, the claustrum could act as a N320 component than non-Aha-answers; Qiu et al., 2006), the intro- monitoring center for mind-wandering, allowing the subject's mind to duction of another variable (i.e. the comprehension of the correct an- drift off during the first reorganization of the available information, let- swer) led to another (more specific) conclusion. The fact that non- ting the correct answer arise from its subconscious representation. comprehended answers also elicited a larger N320 suggests that this Finally, other regions playing a major role in insight processing are lo- component could be a marker of the cognitive conflict that occurs cated in the left prefrontal lobe (i.e., IFG and MFG), precuneus and inferior when new ways of thinking are employed to overcome the cognitive occipital regions (Qiu et al., 2010). Interestingly, a distinction between impasse. “preparation” for insight problem-solving and actual insight processing has been proposed (Tian et al., 2011), suggesting a tighter link between 5.2.2. Later negative components the first processing step and IFG-MFG, left temporal lobe and bilateral Moving toward later time windows, deflections with different laten- claustrum activity. This pool of regions has been suggested as a separate cies have also been correlated with insight problem-solving in several network whose activity precedes the activation of structures in the right studies (Fig. 3). In a pair of studies by Qiu and colleagues using the Chi- temporal lobe responsible for the actual Aha! experience (Tian et al., nese logogriphs (Qiu et al., 2008; Wang et al., 2009), late negative de- 2011; Zhao, Zhou, Xu, Fan, & Han, 2014b). Such hemispheric segregation flections in the 1500–2000 ms and 2000–2500 ms latency ranges – or interplay if one looks at the overall process – is the most popular the- were related to insight problem-solving. In their first study, these com- ory on the neural basis of creativity and divergent thinking in general ponents showed distinct activations over left frontal scalp regions with (Beaty et al., 2015; Beeman & Bowden, 2000; Jung-Beeman et al., 2004). generators located, respectively, in the ACC and in the posterior cingu- Additionally, recent studies suggest a tight link between individual cogni- late cortex (PCC). The PCC involvement (an area that has been related tive profile and hemispheric differences in the spontaneous electrical to the cognitive processing of emotions; Aoki, Cortese, & Tansella, (Gotts et al., 2013) and BOLD-related connectivity patterns 2015) is interesting because it could be a marker of the feeling of (Santarnecchi, Tatti, Rossi, Serino, & Rossi, 2015), suggesting the impor- warmth that follows an Aha! moment. In contrast, the dipole analysis tance of future investigations aimed at uncovering the relationship be- performed in the second study (Wang et al., 2009) found the source of tween insight and interhemispheric dynamics (see the next paragraph). this component in the parahippocampal gyrus and the superior frontal As we have highlighted, insight network shared many areas with sa- gyrus, two areas that respectively involved in forming novel and effec- lience, executive control, fluid intelligence and creativity networks. How- tive associations and in the resolution of conflicts (Luo & Niki, 2003a). ever, the relative rarity of fMRI studies (thirteen in literature) and the An explanation to these non-consistent findings could be due to differ- variety of task and protocols used, (Table S1, Fig. 1), limits the validity ent protocols used in the experiments: a learning-testing model in the of these results. More standardized experimental protocols are needed, first study with true- and false-matching logogriphs and a simple solv- with a larger sample of fMRI investigations for each type of insight task, ing paradigm in the second one. in order to reveal the real insight neural substrates and the role of the Using a different task, Luo et al. (2011) compared structural similar- overlaps with others networks. ity logogriphs (STSL) with surface similarity logogriphs (SUSL) and 114 G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 found a negative peak in the time range of 1100–1300 ms in posterior 5.3. Hemispheric asymmetry and insightful processing scalp regions for STSL (Luo et al., 2011). This component may also reflect a delayed N400-like component taking into account their similar topo- A large part of neuroimaging and neurophysiological studies con- graphical distribution, highlighting the need to control for late solvers. verge to the notion that hemispheric asymmetry as the fundamental structural and a potential functional basis for insight problem-solving (Dietrich & Kanso, 2010), with a prominent role of the right hemisphere 5.2.3. Positive components around 300 ms (Bowden et al., 2005; Kounios & Beeman, 2014). In particular, a critical Positive components have been typically interpreted as a mark of role has been attributed to the right anterior superior temporal gyrus the creation of novel associations, in the frame of the “context closure” because of its involvement in finding distant sematic relationships be- model (Picton, 1992). The P300 component, one of the most studied tween words and, in general, in coarse sematic coding (Bowden & ERP waveforms, has been associated with online WM updating and rec- Jung-Beeman, 2007). However, our analysis of the literature does not ollection of information stored in long-term memory. Its latency varies support this view, with a less striking lateralization of brain oscillatory between 300 and 600 ms from stimulus onset based on the duration activity (Fig. 4) and/or fMRI activations during both mental preparation of stimulus classification, and its amplitude seems to reflect attentional and insightful events (Fig. 5; i.e. from our ALE maps only middle tempo- load (Wilson, Harkrider, & King, 2012). P300-like components have ral gyrus and precuneus show a unilateral activation pattern, Fig. 2). been the focus of attention of several ERP studies on insight problem- Rather, in the “pure insight” studies (Fig. 4), most of the activations solving (Fig. 3). Qiu et al. (2008) showed that successfully completed are widespread and bilateral, with bilaterality supported also by logogriphs were correlated with a larger positive deflection perturbation-based investigation using rTMS (Giovannelli et al., 2010), (P200–600) than unsuccessfully completed logogriphs, with generators as well as by most of NiBS studies (see below and Fig. 6). This is partic- localized to the left superior temporal gyrus and parieto-temporo- ularly relevant, since NiBS is the only approach that can provide causal, occipital cortex (Qiu et al., 2008). These areas, commonly associated rather than correlational, information. Only few oscillatory patterns (i.e. with the creation of associations (Luo et al., 2003b), could be interpreted γ activity burst in temporal and frontal lobes, α band activity in the pa- as the initial attempt to inhibit the superficial meaning of the logogriphs rietal lobe, as well as θ in the temporal lobe) are specific to the right and to find deeper elements useful for their solution. hemisphere (Fig. 4; Jung-Beeman et al., 2004; Danko et al., 2003; Nevertheless, given the relatively short latency of these components, Sheth et al., 2009). In terms of left hemispheric activation, we only they have been also related to the process of breaking mental set. Wang found evidence of a link between insight and increased coherence in β et al. (2009) compared insight with routine problems and found that in- and α frequency bands in the left parietal and temporal lobes during insight problem-solving was associated with a more positive ERP deflec- sight problem-solving (Danko et al., 2003; Razumnikova, 2007). The tion (P300–800) over parieto-occipital scalp regions than routine same bilateral pattern seems to be present also during the mental prep- problems, with generators localized in the ACC and parahippocampal aration (Kounios et al., 2006), with results showing a widespread, bilat- gyrus. Another change in the adopted reference state led to an addition- eral decrease in α power (Fig. 5). al finding: insight solutions showed a greater positive deflection As for resting-state EEG activity, an interesting pattern arises when (P400–600) when compared to search solutions (analytical problem- the activity of subjects unaware about the type of task they had to solving). Intriguingly, this activity was shown to generate from the fusi- solve is taken into account (Fig. 5). Indeed, lateralized activations corre- form gyrus, an area implied in perceptual processing (Zhang, Tian, Wu, lated with higher insight ability: spectral power in the β and γ frequen- Liao, & Qiu, 2011). cy bands are increased in the right frontal lobe, whereas an increase in α Finally, the paradigm (Chinese character-generation task) adopted and γ power are reported in, respectively, the left frontal and left tem- by Zhao et al. (2011) showed a larger positive deflection (P500–700) poral lobe regions (Kounios et al., 2008). This is interesting considering in the “breaking mental set” condition (i.e. participants had to use a dif- how a link between brain functional asymmetry at rest and cognition in ferent method to create a new Chinese character; Zhao et al., 2011)than humans has been recently experimentally supported (Santarnecchi in “repetition condition” (i.e. participants generate a new Chinese char- et al., 2015b), as well as many of the lateralized resting-state oscillatory acter using the preceding method), providing further evidence for later patterns at hand display a flipped pattern respect to what has been de- P300 components as a marker of breaking the mental set. scribed during actual insight problem-solving (e.g. γ burst in the right temporal lobe). 5.2.4. Later positive components In conclusion, even though the EEG and fMRI literature is not ho- Further evidence suggests that the early positive deflections mogenous and only a handful of studies are available, the results seem outlined above are followed by additional later positive deflections to not support a complete dominance of the right hemisphere during (Fig. 3). These have been generally interpreted as markers of the crea- Eureka moments. It seems more likely that right activations are more tion of novel associations and with rehearsal/retention operations in pronounced specifically during resting-state activity of high insight sub- WM. More positive ERP deflections related to insight solutions were ject, whereas insight problem-solving is more related to a widespread found by Wang et al. (2009) 1200–1500 ms after the onset of the stim- synergistic activation of bilateral areas, with specific brain oscillations uli, and in the 640–780 ms time window in a later study by the same playing a major role in the right hemisphere (i.e. γ burst in temporal group (Zhang et al., 2011). Dipole analysis of the 1200–1500 ms peak and frontal lobe). comparing insight and routine problems (Wang et al., 2009)showeda generator in the parahippocampal and in the superior frontal gyri, 5.4. Electric signatures of an insightful mind while the generator for the 640–780 ms peak (comparing insight solutions and search solutions) was localized in the right superior temporal Despite the use of similar tasks, an agreement about the role of dif- gyrus. Both results stress the role of parahippocampal gyrus in the gen- ferent brain oscillations during insight does not appear to be present eration of connections between elements and support the idea of the to date (Fig. 4). However, a major role for α (and possibly γ) activity superior frontal gyrus being related to conflict resolution. Similarly, seems to emerge, with the most consistent findings suggesting a reduc- Luo et al. (2011) found another positive slow going wave between tion of α power and coherence in the frontal lobe, coupled with an in- 900 and 1700 ms in SUSL and STSL trials, namely when the subjects crease in α power in the right parietal lobe during insight processing try to find a new character after understanding or breaking the surface (Fig. 4). As for ERPs, overall the increase of the N400 amplitude seems meaning of the logogriph. Therefore, it seems reasonable that later pos- to be one of the most replicated findings across all the literature, as itive components could play a role in the processes of forming new as- well as the positive deflection P300 (Fig. 3). There is a subtle concor- sociations across pieces of information. dance in the ERPs correlates of insight in the various studies, but their G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 115 localization abilities as well as the mechanistic information they convey state-dependent effects of alternating current stimulation (tACS) are are far to be exhaustive. Certainly, activity recorded along the midline exploited (Feurra et al., 2011a, 2011b; Santarnecchi et al., 2013). As seems to be the most widely reported. However, fMRI data show a dif- for real brain lesions, a study by Reverberi and colleagues (Reverberi, ferent pattern, with regions from different brain lobes taking part in Toraldo, D'Agostini, & Skrap, 2005)confirmed the causal role of lateral the Eureka moment (Fig. 2). Overall, it seems that regions involved in at- frontal cortex in insight problem-solving ability: patients with lateral tention/salience play a pivotal role during insight processes, with a pos- frontal damage were more successful in solving difficult matchstick ar- sibly underestimated role for the claustrum. The only region that ithmetic problems with respect to healthy subjects, while medial broadly aligns with the existing literature – even though related to a se- frontal-lesion patients shown less pronounced difference in perfor- ries of studies performed by the same group – is the right temporal pole, mance. Unfortunately, this constitutes to date the only clinical evalua- where increase in both BOLD activation and EEG activity – represented tion of brain lesions’ impact on insight ability. by focal γ bursts – have been reported specifically during insight processing but not during non-insightfully generated answers (Fig. 7; Jung-Beeman et al., 2004). 5.6. Caveats and future directions

5.5. Insight and NiBS All the experiments considered in this review only include right- handed subjects. This could be an important bias for the identification Only a handful of studies assessed the effects of tES or TMS on the Eu- of lateralized and diffuse patterns of activity. The variability in the reka moment, with tES investigations limited to tDCS (Fig. 6, Tables 2 tasks used and the low number of available studies make it difficult to and S4). Evidence converges on the role of the temporal lobes, where get a clear picture of the neurobiological underpinnings of the Eureka an increase in cortical excitability in the right hemisphere induced by moment. While the review at hand probably represents the most com- anodal tDCS (coupled with a decrease in the contralateral one) seems plete quantitative meta-analysis on insight available to date, a able to improve insight-related performance. As for prefrontal stimula- parcellation of our results in task-specific ALE maps as well as process- tion, only stimulation of the left hemisphere lead to performance en- specific EEG evoked-activity would be ideal. More carefully designed in- hancement in both studies available to date (Cerruti & Schlaug, 2009; vestigations are needed and collaborative, multicenter efforts should be Metuki et al., 2012), suggesting a dominant role of this region, a finding pursued. which is in line with our ALE map results. The only rTMS study available As for NiBS, investigations testing the feasibility of frequency- so far (Giovannelli et al., 2010) reports similar enhancing effects for specific modulation of insight problem-solving are needed, for instance stimulation of right and left parietal cortex. However, this result could using tACS to test the impact of transcranially induced oscillatory poten- be due to the specific task used during rTMS (i.e. binarized images tials on the EEG signatures of insight summarized in Figs. 4 and 5.The task), which allows for the assessment of a very specificsubcomponent modulation of γ and α oscillations while subjects are solving insight of insight processing (perceptual ability) and probably not insight tasks might provide a causal evidence of their relevance. In a near fu- problem-solving per se. ture, closed-loop protocols, where electrical or magnetic stimulation is It is important to note that, despite the apparent effectiveness of pre- delivered according to specific patterns of ongoing EEG activity, could frontal lobe stimulation during insight processes, the variability in elec- probably shed additional light on the physiological mechanism under- trode montages, especially related to the positioning of the so called pinning insight problem-solving and α/γ activity roles in its dynamics. “return” electrode, suggests valid concerns about the specificroleattrib- uted to the “target” region. With tDCS being a bipolar stimulation where both electrodes deliver an equal intensity but opposite polarity electri- 6. Conclusion cal stimulation, it is misleading to attribute null effects to the electrode placed as “return” (i.e. cathode during so-called anodal tDCS, and vice Despite the many neuroimaging and electrophysiological investiga- versa; Santarnecchi et al., 2015a). Therefore, tES evidence collected so tions conducted to assess the neural basis of insight during the last far might suffer from a generalized lack of focality: anodal stimulation twenty years, the findings are more controversial than expected. Evi- over the left prefrontal lobe while applying the cathode over the right dence about specific insight-related oscillatory patterns are emerging, prefrontal cortex might induce significant effects due to (i) increased such as the role of α band in frontal and parietal lobe (Fig. 7). The anal- cortical excitability of the former, (ii) decreased excitability of the latter ysis of resting-state activity and mental preparation, although still in its (with potential cascade effects over interhemispheric inhibitory pro- infancy, reveals interesting findings such as segregation of right hemi- cesses on top of local decrease in activation), and (iii) the combination sphere activity in the frontal lobe, as well as a widespread decrement of both effects. Careful selection of both active and return electrodes po- of α power (Fig. 5). Additionally, the conception of right hemispheric sitions is a crucial requirement for future studies on insight, with solu- dominance during insight problem-solving does not appear to be con- tions as multifocal/multielectrode stimulation to be considered firmed by our meta-analysis of fMRI activations (Fig. 2), although it par- (Ruffini, Fox, Ripolles, Miranda, & Pascual-Leone, 2014). tially fits with electrophysiological data (Fig. 4). Testing insight abilities Clearly, additional neuromodulatory experiments are needed in in subjects with full right hemispheric dominance might be helpful in order to enlighten the causal role of specific brain region and their oscil- this sense. Furthermore, despite showing promisingly coherent results, latory patterns during the Eureka moment. As shown by its exponential neuromodulatory studies have been focused mainly on tDCS over two application for cognitive enhancement and rehabilitation in the last ten brain regions. More consistent findings seem to be related to stimula- years (Filmer, Dux, & Mattingley, 2014; Nitsche & Paulus, 2011; Rossi & tion of the right temporal lobe – in accordance with converging EEG Rossini, 2004; Santarnecchi et al., 2015a), tES should be thought of as a and neuroimaging evidence – and the prefrontal cortex, despite the lat- double purpose tool: while the “modulation” of local and distant brain eralization not being concordant across the studies (Fig. 7). Finally, con- activity has been considered the reason for the reintroduction of tES in sidering the large variety of insight tasks being used despite their lack of modern neuroscience (e.g., Liew, Santarnecchi, Buch, & Cohen, 2014), ecological or statistical properties, the validation of new insight tasks its potential as a tool for the investigation of the causal role of specific able to fully capture the individual variability in covert cognitive pro- brain regions in a given network/function should not necessarily come cesses behind a successful Aha! moment should constitute a priority as a secondary feature. As observed for TMS in the context of the “virtual for the scientificcommunity. lesion” approach (Pascual-Leone, Walsh, & Rothwell, 2000), tES might Understanding the neurophysiology underpinning the Eureka mo- help establishing specific contribution of a given network, especially in ment, a fundamental manifesto of the human mind, is a central neuro- terms of brain oscillatory correlates of Aha! when frequency- and scientific challenge that should not be postponed further. 116 G. Sprugnoli et al. / Intelligence 62 (2017) 99–118

Financial disclosures Bowden, E. M., & Jung-Beeman, M. (1998). Getting the right idea: Semantic activation in the right hemisphere may help solve insight problems. Psychological Science, 9, 435–440. All authors report no conflicts of interest. Bowden,E.M.,&Jung-Beeman,M.(2003).Aha! insight experience correlates with solution activation in the right hemisphere. Psychonomic Bulletin & Review, 10, 730–737. Acknowledgements Bowden, E. M., & Jung-Beeman, M. (2007). Methods for investigating the neural components of insight. Methods, 42,87–99. Bowden, E. M., Jung-Beeman, M., Fleck, J., & Kounios, J. (2005). New approaches to Emiliano Santarnecchi and Alvaro Pascual-Leone are supported by demystifying insight. Trends in Cognitive Sciences, 9,322–328. the Intelligence Advanced Research Projects Activity (IARPA) via Bressler, S. L., & Menon, V. (2010). Large-scale brain networks in cognition: Emerging methods and principles. Trends in Cognitive Sciences, 14,277–290. contract #: 2014-13121700007 issued to Honeywell International in Carson, S. H., Peterson, J. B., & Higgins, D. M. (2003). Decreased latent inhibition is associ- the context of the Strengthening Human Adaptive Reasoning and ated with increased creative achievement in high-functioning individuals. Journal of – Problem-Solving (SHARP) program. Dr. Pascual-Leone is further Personality and Social Psychology, 85,499 506. Cerruti, C., & Schlaug, G. (2009). Anodal transcranial direct current stimulation of the pre- supported by the Berenson-Allen Foundation, the Sidney R. Baer Jr. frontal cortex enhances complex verbal associative thought. Journal of Cognitive Foundation, grants from the National Institutes of Health (R01 Neuroscience, 21, 1980–1987. HD069776, R01 NS073601, R21 MH099196, R21 NS082870, R21 Chein, J. M., Weisberg, R. W., Streeter, N. L., & Kwok, S. (2010). Working memory and insight in the nine-dot problem. Memory & Cognition, 38,883–892. NS085491, R21 HD07616), and Harvard Catalyst | The Harvard Clinical Chi, R. P., & Snyder, A. W. (2011). Facilitate insight by non-invasive brain stimulation. PLoS and Translational Science Center (NCRR and the NCATS NIH, UL1 One, 6,e16655. RR025758). The content of this paper is solely the responsibility of the Chi, R. P., & Snyder, A. W. (2012). Brain stimulation enables the solution of an inherently fi – fi dif cult problem. Neuroscience Letters, 515,121 124. authors and does not necessarily represent the of cial views of Harvard Chu, Y., & MacGregor, J. N. (2011). Human performance on insight problem solving: A re- Catalyst, Harvard University and its affiliated academic health care cen- view. The Journal of Problem Solving, 3,119–150. ters, the National Institutes of Health, the Sidney R. Baer Jr. Foundation, Chudersky, A. (2014). How well can storage capacity, executive control, and fluid reason- – the United States Government, the Office of the Director of National In- ing explain insight problem solving. Intelligence, 46,258 270. Colom, R., Jung, R. E., & Haier, R. J. (2007). General intelligence and memory span: Evi- telligence, or the Intelligence Advanced Research Projects Activity. dence for a common neuroanatomic framework. Cognitive Neuropsychology, 24, 867–878. — Appendix A. Supplementary data Craig, A. D. (2009). How do you feel Now? The anterior insula and human awareness. Nature Reviews Neuroscience, 10,59–70. Dandan, T., Haixue, Z., Wenfu, L., Wenjing, Y., Jiang, Q., & Qinglin, Z. (2013). Brain activity Supplementary data to this article can be found online at http://dx. in using heuristic prototype to solve insightful problems. Behavioural Brain Research, – doi.org/10.1016/j.intell.2017.03.004. 253,139 144. Danko, S. G., Starchenko, M. G., & Bechtereva, N. P. (2003). EEG local and spatial synchronization during a test on the insight strategy of solving creative verbal tasks. Human Physiology, 29,502–504. References Dietrich, A., & Kanso, R. (2010). A review of EEG, ERP, and neuroimaging studies of creativity and insight. Psychological Bulletin, 136,822–848. Abraham, A., & Windmann, S. (2007). Creative cognition: The diverse operations and the Eickhoff, S. B., Bzdok, D., Laird, A. R., Kurth, F., & Fox, P. T. (2012). Activation likelihood es- prospect of applying a cognitive neuroscience perspective. Methods, 42,38–48. timation meta-analysis revisited. NeuroImage, 59,2349–2361. Ackermann, H., & Riecker, A. (2004). The contribution of the insula to motor aspects of Eickhoff, S. B., Heim, S., Zilles, K., & Amunts, K. (2009b). A systems perspective on the ef- speech production: A review and a hypothesis. Brain and Language, 89,320–328. fective connectivity of overt speech production. Philosophical Transactions. Series A, Adank, P. (2012). The neural bases of difficult speech comprehension and speech produc- Mathematical, Physical, and Engineering Sciences, 367,2399–2421. tion: Two Activation Likelihood Estimation (ALE) meta-analyses. Brain and Language, Eickhoff, S. B., Laird, A. R., Grefkes, C., Wang, L. E., Zilles, K., & Fox, P. T. (2009a). 122,42–54. Coordinate-based activation likelihood estimation meta-analysis of neuroimaging Aldous, C. R. (2007). Creativity, problem solving and innovative science: Insights from data: A random-effects approach based on empirical estimates of spatial uncertainty. history, cognitive psychology and neuroscience. International Education Journal, 8, Human Brain Mapping, 30,2907–2926. 176–186. Estrada, C., Joung, M., & Isen, A. M. (1994). Positive affect influences creative problem Amabile, T. M., Barsade, S. G., Mueller, J. S., & Staw, B. M. (2005). Affect and creativity at solving and reported source of practice satisfaction in physicians. Motivation and work. Administrative Science Quarterly, 50,367–403. Emotion, 18,285–299. Aoki, Y., Cortese, S., & Tansella, M. (2015). Neural bases of atypical emotional face process- Fall,S.,Querne,L.,LeMoing,A.G.,&Berquin,P.(2015).Individual differences in sub- ing in autism: A meta-analysis of fMRI studies. The World Journal of Biological cortical microstructure organization reflect reaction time performances during a Psychiatry, 16,291–300. flanker task: A diffusion tensor imaging study in children with and without Ardila, A., Bernal, B., & Rosselli, M. (2014). Participation of the insula in language revisited: ADHD. Psychiatry Research, 233,50–56. A meta-analytic connectivity study. Journal of Neurolinguistics, 29,31–41. Federmeier, K. D., Kirson, D. A., Moreno, E. M., & Kutas, M. (2001). Effects of transient, mild Augustine, J. R. (1996). Circuitry and functional aspects of the insular lobe in primates in- mood states on semantic memory organization and use: An event-related potential cluding humans. Brain Research. Brain Research Reviews, 22,229–244. investigation in humans. Neuroscience Letters, 305,149–152. Aziz-Zadeh, L., Kaplan, J. T., & Iacoboni, M. (2009). “Aha!”: The neural correlates of verbal Feurra, M., Bianco, G., Santarnecchi, E., Del, T. M., Rossi, A., & Rossi, S. (2011a). Frequency- insight solutions. Human Brain Mapping, 30,908–916. dependent tuning of the human motor system induced by transcranial oscillatory po- Barbey, A. K., Colom, R., & Grafman, J. (2013). Architecture of cognitive flexibility revealed tentials. The Journal of Neuroscience, 31,12165–12170. by lesion mapping. NeuroImage, 82,547–554. Feurra, M., Paulus, W., Walsh, V., & Kanai, R. (2011b). Frequency specific modulation of Beaty, R. E., Benedek, M., Kaufman, S. B., & Silvia, P. J. (2015). Default and executive net- human somatosensory cortex. Frontiers in Psychology, 2,13. work coupling supports creative idea production. Science Reports, 5,10964. Filmer, H. L., Dux, P. E., & Mattingley, J. B. (2014). Applications of transcranial direct cur- Beeman, M. J., & Bowden, E. M. (2000). The right hemisphere maintains solution-related rent stimulation for understanding brain function. Trends in Neurosciences, 37, activation for yet-to-be-solved problems. Memory & Cognition, 28,1231–1241. 742–753. Benn, Y., Webb, T. L., Chang, B. P., Sun, Y. H., Wilkinson, I. D., & Farrow, T. F. (2014). The Fink, A., Koschutnig, K., Benedek, M., Reishofer, G., Ischebeck, A., Weiss, E. M., et al. (2012). neural basis of monitoring goal progress. Frontiers in Human Neuroscience, 8,688. Stimulating creativity via the exposure to other people's ideas. Human Brain Mapping, Bermejo, M. R., Castejon, J. L., & Sternberg, R. J. (1996). Insight in children with high intel- 33,2603–2610. ligence level. FAISCA, 4,85–94. Fleck, J. I. (2008). Working memory demands in insight versus analytic problem solving. deBettencourt, M. T., Cohen, J. D., Lee, R. F., Norman, K. A., & Turk-Browne, N. B. (2015). Journal of Cognitive Psychotherapy, 20,139–176. Closed-loop training of attention with real-time brain imaging. Nature Neuroscience, Gibson, C., Folley, B. S., & Park, S. (2009). Enhanced divergent thinking and creativity in 18,470–475. musicians: A behavioral and near-infrared spectroscopy study. Brain and Cognition, Blank, S. C., Scott, S. K., Murphy, K., Warburton, E., & Wise, R. J. (2002). Speech production: 69,162–169. Wernicke, Broca and beyond. Brain, 125,1829–1838. Gilhooly, K. J., & Fioratou, E. (2009). Executive functions in insight versus non-insight Boccia, M., Piccardi, L., Palermo, L., Nori, R., & Palmiero, M. (2015). Where do bright ideas problem solving. Thinking and Reasoning, 15,355–376. occur in our brain? Meta-analytic evidence from neuroimaging studies of domain- Giovannelli, F., Silingardi, D., Borgheresi, A., Feurra, M., Amati, G., Pizzorusso, T., et al. specific creativity. Frontiers in Psychology, 6,1195. (2010). Involvement of the parietal cortex in perceptual learning (Eureka effect): Bogels, S., Barr, D. J., Garrod, S., & Kessler, K. (2015). Conversational interaction in the An interference approach using rTMS. Neuropsychologia, 48,1807–1812. scanner: Mentalizing during language processing as revealed by MEG. Cerebral Goll, Y., Atlan, G., & Citri, A. (2015). Attention: The claustrum. Trends in Neurosciences, 38, Cortex, 25,3219–3234. 486–495. Boroojerdi, B., Phipps, M., Kopylev, L., Wharton, C. M., Cohen, L. G., & Grafman, J. (2001). Gotts,S.J.,Jo,H.J.,Wallace,G.L.,Saad,Z.S.,Cox,R.W.,&Martin,A.(2013).Two dis- Enhancing analogic reasoning with rTMS over the left prefrontal cortex. Neurology, tinct forms of functional lateralization in the human brain. Proceedings of the 56,526–528. National Academy of Science of the United States of America, 110, E3435–E3444. G. Sprugnoli et al. / Intelligence 62 (2017) 99–118 117

Guilford, J. P. (1967). The nature of human intelligence. New York, NY: McGraw-Hill. Minami, T., Noritake, Y., & Nakauchi, S. (2014). Decreased beta-band activity is correlated Hao, X., Cui, S., Li, W., Yang, W., Qiu, J., & Zhang, Q. (2013). Enhancing insight in scientific with disambiguation of hidden figures. Neuropsychologia, 56,9–16. problem solving by highlighting the functional features of prototypes: An fMRI study. Mion, M., Patterson, K., Acosta-Cabronero, J., Pengas, G., Izquierdo-Garcia, D., Hong, Y. T., Brain Research, 1534,46–54. et al. (2010). What the left and right anterior fusiform gyri tell us about semantic Isen, A. M., Daubman, K. A., & Nowicki, G. P. (1987). Positive affect facilitates creative memory. Brain, 133, 3256–3268. problem solving. Journal of Personality and Social Psychology, 52,1122–1131. Moore, T., & Fallah, M. (2004). Microstimulation of the frontal eye field and its effects on Jung, R. E., & Haier, R. J. (2007). The Parieto-Frontal Integration Theory (P-FIT) of intelli- covert spatial attention. Journal of Neurophysiology, 91,152–162. gence: Converging neuroimaging evidence. The Behavioral and Brain Sciences, 30, Mourgues, C. V., Preiss, D. D., & Grigorenko, E. L. (2014). Reading skills, creativity, and in- 135–154. sight: Exploring the connections. Spanish Journal of Psychology, 17, E58. Jung-Beeman, M., Bowden, E. M., Haberman, J., Frymiare, J. L., Arambel-Liu, S., Greenblatt, Mumford, M. D., & Whetzel, D. L. (1996). Insight, creativity, and cognition: On Sternberg R., et al. (2004). Neural activity when people solve verbal problems with insight. PLoS and Davidson's the nature of insight. Creativity Research Journal, 9(1), 103–107. Biology, 2, E97. Murray, M. A., & Byrne, R. M. J. (2005). Attention and working memory in insight Knoblich, G., Ohlsson, S., Haider, H., & Rhenius, D. (1999). Constraint relaxation and chunk problem-solving. In B. Bara, L. Barsalou, & M. Bucciareli (Eds.), Proceedings of 27th an- decomposition in insight problem solving. Journal of Experimental Psychology: nual meeting of Cognitive Science Society (pp. 1571–1576). Learning, Memory, and Cognition, 25,1534–1555. Nieuwenhuys, R. (2012). The insular cortex: A review. Progress in Brain Research, 195, Kohler, W. (1925). The mentality of apes. London: Routledge & Kegan Paul. 123–163. Kosslyn, S. M., & Thompson, W. L. (2003). When is early visual cortex activated during vi- Nitsche, M. A., & Paulus, W. (2011). Transcranial direct current stimulation—Update 2011. sual mental imagery? Psychological Bulletin, 129,723–746. Restorative Neurology and Neuroscience, 29,463–492. Koubeissi, M. Z., Bartolomei, F., Beltagy, A., & Picard, F. (2014). Electrical stimulation of a Novick, L. R., & Sherman, S. J. (2003). On the nature of insight solutions: Evidence from small brain area reversibly disrupts consciousness. Epilepsy & Behavior, 37,32–35. skill differences in anagram solution. The Quarterly Journal of Experimental Kounios, J. (1996). On the continuity of thought and the representation of knowledge: Psychology. A, 56,351–382. Electrophysiological and behavioral time-course measures reveal levels of structure Oh, A., Duerden, E. G., & Pang, E. W. (2014). The role of the insula in speech and language in semantic memory. Psychonomic Bulletin & Review, 3,265–286. processing. Brain and Language, 135,96–103. Kounios, J., & Beeman, M. (2014). The cognitive neuroscience of insight. Annual Review of Ostafin, B. D., & Kassman, K. T. (2012). Stepping out of history: Mindfulness improves in- Psychology, 65,71–93. sight problem solving. Consciousness and Cognition, 21, 1031–1036. Kounios, J., Fleck, J. I., Green, D. L., Payne, L., Stevenson, J. L., Bowden, E. M., et al. (2008). Pascual-Leone, A., Walsh, V., & Rothwell, J. (2000). Transcranial magnetic stimulation in The origins of insight in resting-state brain activity. Neuropsychologia, 46,281–291. cognitive neuroscience—Virtual lesion, chronometry, and functional connectivity. Kounios, J., Frymiare, J. L., Bowden, E. M., Fleck, J. I., Subramaniam, K., Parrish, T. B., et al. Current Opinion in Neurobiology, 10,232–237. (2006). The prepared mind: Neural activity prior to problem presentation predicts Paulewicz, B., Chudersky, A., & Necka, E. (2007). Insight problem solving, fluid intelli- subsequent solution by sudden insight. Psychological Science, 17,882–890. gence, and executive control: A structural equation modeling approach. In S. Kroger, J. K., Sabb, F. W., Fales, C. L., Bookheimer, S. Y., Cohen, M. S., & Holyoak, K. J. (2002). Vosniadou, D. Kayser, & A. Protopapas (Eds.), Proceedings of the 2nd European cogni- Recruitment of anterior dorsolateral prefrontal cortex in human reasoning: A para- tive science conference (pp. 586–591). metric study of relational complexity. Cerebral Cortex, 12,477–485. Picton, T. W. (1992). The P300 wave of the human event-related potential. Journal of Lang, S., Kanngieser, N., Jaskowski, P., Haider, H., Rose, M., & Verleger, R. (2006). Precur- Clinical Neurophysiology, 9,456–479. sors of insight in event-related brain potentials. Journal of Cognitive Neuroscience, Qiu, J., Li, H., Jou, J., Liu, J., Luo, Y., Feng, T., et al. (2010). Neural correlates of the “Aha” ex- 18,2152–2166. periences: Evidence from an fMRI study of insight problem solving. Cortex, 46, Lau, E. F., Phillips, C., & Poeppel, D. (2008). A cortical network for semantics: 397–403. (De)constructing the N400. Nature Reviews Neuroscience, 9,920–933. Qiu, J., Li, H., Luo, Y., Chen, A., Zhang, F., Zhang, J., et al. (2006). Brain mechanism of cog- Liew, S. L., Santarnecchi, E., Buch, E. R., & Cohen, L. G. (2014). Non-invasive brain stimula- nitive conflict in a guessing Chinese logogriph task. Neuroreport, 17,679–682. tion in neurorehabilitation: Local and distant effects for motor recovery. Frontiers in Qiu, J., Li, H., Yang, D., Luo, Y., Li, Y., Wu, Z., et al. (2008). The neural basis of insight prob- Human Neuroscience, 8,378. lem solving: An event-related potential study. Brain and Cognition, 68,100–106. Luo, J., Li, W., Fink, A., Jia, L., Xiao, X., Qiu, J., et al. (2011). The time course of breaking men- Razumnikova, O. M. (2007). Creativity related cortex activity in the remote associates tal sets and forming novel associations in insight-like problem solving: An ERP inves- task. Brain Research Bulletin, 73,96–102. tigation. Experimental Brain Research, 212,583–591. Reverberi, C., Toraldo, A., D'Agostini, S., & Skrap, M. (2005). Better without (lateral) frontal Luo, J., Li, W., Qiu, J., Wei, D., Liu, Y., & Zhang, Q. (2013). Neural basis of scientificinnova- cortex? Insight problems solved by frontal patients. Brain, 128,2882–2890. tion induced by heuristic prototype. PLoS One, 8,e49231. Rossi, S., & Rossini, P. M. (2004). TMS in cognitive plasticity and the potential for rehabil- Luo, J., & Niki, K. (2003a). Function of hippocampus in “insight” of problem solving. itation. Trends in Cognitive Sciences, 8,273–279. Hippocampus, 13,316–323. Roth, J. K., Serences, J. T., & Courtney, S. M. (2006). Neural system for controlling the con- Luo, J., Niki, K., & Phillips, S. (2004a). Neural correlates of the ‘Aha! reaction’. Neuroreport, tents of object working memory in humans. Cerebral Cortex, 16,1595–1603. 15,2013–2017. Rowe, G., Hirsh, J. B., & Anderson, A. K. (2007). Positive affect increases the breadth of at- Luo, J., Niki, K., & Phillips, S. (2004b). The function of the anterior cingulate cortex (ACC) in tentional selection. Proceedings of the National Academy of Science of the United States the insightful solving of puzzles: The ACC is activated less when the structure of the of America, 104,383–388. puzzle is known. Journal of Psychology in Chinese Societies, 5,195–213. Ruffini, G., Fox, M. D., Ripolles, O., Miranda, P. C., & Pascual-Leone, A. (2014). Optimization Luo, Q., Perry, C., Peng, D., Jin, Z., Xu, D., Ding, G., et al. (2003b). The neural substrate of of multifocal transcranial current stimulation for weighted cortical pattern targeting analogical reasoning: An fMRI study. Brain Research. Cognitive Brain Research, 17, from realistic modeling of electric fields. NeuroImage, 89,216–225. 527–534. Salvi, C., Bricolo, E., Franconeri, S. L., Kounios, J., & Beeman, M. (2015). Sudden insight is Lustenberger, C., Boyle, M. R., Foulser, A. A., Mellin, J. M., & Frohlich, F. (2015). Functional associated with shutting out visual inputs. Psychonomic Bulletin & Review, 22, role of frontal alpha oscillations in creativity. Cortex, 67,74–82. 1814–1819. MacGregor, J. N., & Cunningham, J. B. (2008). Rebus Puzzles as insight problems. Behavior Salvi, C., Bricolo, E., Kounios, J., Bowden, M., & Beeman, M. (2016). Insight solutions are Research Methods, 40,263–268. correct more often than analytic solutions. Thinking & Reasoning, 22,443–460. MacGregor, J. N., Ormerod, T. C., & Chronicle, E. P. (2001). Information processing and in- Sandkühler, S., & Bhattacharya, J. (2008). Deconstructing insight: EEG correlates of insight: A process model of performance on the nine-dot and related problems. Journal sightful problem solving. PLoS One, 3,e1459. of Experimental Psychology. Learning, Memory, and Cognition, 27,176–201. Santarnecchi, E., Brem, A. K., Levenbaum, E., Thompson, T., Kadosh, R. C., & Pascual-Leone, Mai, X. Q., Luo, J., Wu, J. H., & Luo, Y. J. (2004). “Aha!” effects in a guessing riddle task: An A. (2015a). Enhancing cognition using transcranial electrical stimulation. Current event-related potential study. Human Brain Mapping, 22,261–270. Opinion in Behavioural Sciences,171–178. Martindale, C. (1999). The biological basis of creativity. Cambridge, England: Cambridge Santarnecchi, E., D'Arista, S., Egiziano, E., Gardi, C., Petrosino, R., Vatti, G., et al. (2014). In- University Press. teraction between neuroanatomical and psychological changes after mindfulness- Mashal, N., Faust, M., & Hendler, T. (2005). The role of the right hemisphere in processing based training. PLoS One, 9, e108359. nonsalient metaphorical meanings: Application of principal components analysis to Santarnecchi, E., Polizzotto, N. R., Godone, M., Giovannelli, F., Feurra, M., Matzen, L., et al. fMRI data. Neuropsychologia, 43,2084–2100. (2013). Frequency-dependent enhancement of fluid intelligence induced by trans- Mason, R. A., & Just, M. A. (2004). How the brain processes causal inferences in text. cranial oscillatory potentials. Current Biology, 23,1449–1453. Psychological Science, 15,1–7. Santarnecchi, E., Tatti, E., Rossi, S., Serino, V., & Rossi, A. (2015b). Intelligence-related dif- McPherson, W. B., & Holcomb, P. J. (1999). An electrophysiological investigation of se- ferences in the asymmetry of spontaneous cerebral activity. Human Brain Mapping, mantic priming with pictures of real objects. Psychophysiology, 36,53–65. 36,3586–3602. Mednick, S. A. (1962). The associative basis of the creative process. Psychological Review, Schmitt, O., Eipert, P., Kettlitz, R., Lessmann, F., & Wree, A. (2016). Theconnectomeofthe 69,220–232. basal ganglia. Brain Structure & Function, 221,753–814. Mendelsohn, G. A., & Griswold, B. B. (1966). Assessed creative potential, vocabulary level, Schooler, J. W., & Melcher, J. (1997). The ineffability of insight. The creative cognition ap- and sex as predictors of the use of incidental cues in verbal problem solving. Journal of proach. Cambridge (Mass.): MIT, 97–133. Personality and Social Psychology, 4,423–431. Shen, W., Liu, C., Yuan, Y., Zhang, X., & Luo, J. (2013). Temporal dynamics of mental im- Mesulam, M. M., & Mufson, E. J. (1982). Insula of the old world monkey. III: Efferent cor- passes underlying insight-like problem solving. Science China. Life Sciences, 56, tical output and comments on function. Journal of Comparative Neurology, 212,38–52. 284–290. Metuki, N., Sela, T., & Lavidor, M. (2012). Enhancing cognitive control components of in- Sheth, B. R., Sandkühler, S., & Bhattacharya, J. (2009). Posterior beta and anterior gamma sight problems solving by anodal tDCS of the left dorsolateral prefrontal cortex. Brain oscillations predict cognitive insight. Journal of Cognitive Neuroscience, 21,1269–1279. Stimulation, 5,110–115. Simonton, D. K. (2001). The psychology of creativity: historical perspective. Green College Miller, E. K., & Cohen, J. D. (2001). An integrative theory of prefrontal cortex function. Lecture Series on The nature of creativity: History biology, and socio-cultural dimensions. Annual Review of Neuroscience, 24,167–202. University of British Columbia. 118 G. Sprugnoli et al. / Intelligence 62 (2017) 99–118

Smith, J. B., & Alloway, K. D. (2014). Interhemispheric claustral circuits coordinate sensory Wang, T., Zhang, Q., Li, H., Qiu, J., Tu, S., & Yu, C. (2009). The time course of Chinese riddles and motor cortical areas that regulate exploratory behaviors. Frontiers in Systems solving: Evidence from an ERP study. Behavioural Brain Research, 199,278–282. Neuroscience, 8,93. Wegbreit, E., Suzuki, S., Grabowecky, M., Kounios, J., & Beeman, M. (2012). Visual atten- Sridharan, D., Levitin, D. J., & Menon, V. (2008). A critical role for the right fronto-insular tion modulates insight versus analytic solving of verbal problems. Journal of cortex in switching between central-executive and default-mode networks. Problem Solving, 4,94–115. Proceedings of the National Academy of Science of the United States of America, 105, Wharton, C. M., Grafman, J., Flitman, S. S., Hansen, E. K., Brauner, J., Marks, A., et al. (2000). 12569–12574. Toward neuroanatomical models of analogy: A positron emission tomography study Sternberg, R. J., & Davidson, J. E. (1995). The nature of insight (XVIII ed.). Cambridge, MA, of analogical mapping. Cognitive Psychology, 40,173–197. US: The MIT Press. Wieth, M. B., & Zacks, R. T. (2011). Time of day effects on problem solving: When the non- Subramaniam, K. (2008). The behavioral and neural basis for the facilitation of insight optimal is optimal. Thinking and Reasoning, 17,387–401. problem-solving by a positive mood. Northwestern University. Wiley, J., & Jarosz, A. F. (2012). How working memory capacity affects problem solving. Subramaniam, K., Kounios, J., Parrish, T. B., & Jung-Beeman, M. (2009). Abrainmechanism Psychology of Learning and Motivation, 56,185–227. for facilitation of insight by positive affect. Journal of Cognitive Neuroscience, 21, Wilson, M. J., Harkrider, A. W., & King, K. A. (2012). Effects of complexity of visual 415–432. distracters on attention and information processing speed reﬂected in auditory Tang, Q., Li, G., Wang, A., Liu, T., Feng, S., Guo, Z., et al. (2015). A systematic review for the p300. Ear and Hearing, 33,480–488. antidepressant effects of sleep deprivation with repetitive transcranial magnetic Wright, R. D., & Lawrence, M. W. (2008). Orienting of attention. Oxford: Oxford University stimulation. BMC Psychiatry, 15, 282. Press. Tian, F., Tu, S., Qiu, J., Lv, J. Y., Wei, D. T., Su, Y. H., et al. (2011). Neural correlates of mental Wu, L., Knoblich, G., & Luo, J. (2013). The role of chunk tightness and chunk familiarity in preparation for successful insight problem solving. Behavioural Brain Research, 216, problem solving: Evidence from ERPs and fMRI. Human Brain Mapping, 34, 1173–1186. 626–630. Xing, Q., Zhang, J. X., & Zhang, Z. (2012). Event-related potential effetcs associated with Turkeltaub, P. E., Eickhoff, S. B., Laird, A. R., Fox, M., Wiener, M., & Fox, P. (2012). Minimiz- insight problems solving in a Chinese logogriph task. Psychology, 3,65–69. ing within-experiment and within-group effects in Activation Likelihood Estimation Yin, J., Gao, Z., Jin, X., Ding, X., Liang, J., & Shen, M. (2012). The neural mechanisms of meta-analyses. Human Brain Mapping, 33,1–13. percept-memory comparison in visual working memory. Biological Psychology, 90, Van Stockum, C., & DeCaro, M. S. (2013). The path less taken: When working memory ca- 71–79. pacityconstrainsinsight.InM.Knauff,M.Pauen,N.Sebanz,&I.Wachsmuth(Eds.),Pro- Zhang, M., Tian, F., Wu, X., Liao, S., & Qiu, J. (2011). The neural correlates of insight in Chi- ceedings of the 35th annual meeting of the Cognitive Science Society (pp. 3633–3638). nese verbal problems: An event related-potential study. Brain Research Bulletin, 84, Cognitive Science Society: Austin, TX. 210–214. Vartanian, O., & Goel, V. (2005). Task constraints modulate activation in right ventral lat- Zhao, Q., Li, Y., Shang, X., Zhou, Z., & Han, L. (2014a). Uniformity and nonuniformity of eral prefrontal cortex. NeuroImage, 27,927–933. neural activities correlated to different insight problem solving. Neuroscience, 270, Vernet, M., Quentin, R., Chanes, L., Mitsumasu, A., & Valero-Cabre, A. (2014). Frontal eye 203–211. ﬁeld, where art thou? Anatomy, function, and non-invasive manipulation of frontal Zhao,Y.,Tu,S.,Lei,M.,Qiu,J.,Ybarra,O.,&Zhang,Q.(2011).The neural basis of breaking regions involved in eye movements and associated cognitive operations. Frontiers in mental set: An event-related potential study. Experimental Brain Research, 208, Integrative Neuroscience, 8,66. 181–187. Wagner, U., Gais, S., Haider, H., Verleger, R., & Born, J. (2004). Sleep inspires insight. Zhao, Q., Zhou, Z., Xu, H., Chen, S., Xu, F., Fan, W., et al. (2013). Dynamic neural network of Nature, 427,352–355. insight: A functional magnetic resonance imaging study on solving Chinese ‘chengyu’ Wang, H., Callaghan, E., Gooding-Williams, G., McAllister, C., & Kessler, K. (2016). Rhythm riddles. PloS One, 8, e59351. makes the world go round: An MEG-TMS study on the role of right TPJ theta oscilla- Zhao, Q., Zhou, Z., Xu, H., Fan, W., & Han, L. (2014b). Neural pathway in the right hemi- tions in embodied perspective taking. Cortex, 75,68–81. sphere underlies verbal insight problem solving. Neuroscience, 256,334–341.